diff --git "a/data/refined_40k/test_bench/data_lab.json" "b/data/refined_40k/test_bench/data_lab.json" new file mode 100644--- /dev/null +++ "b/data/refined_40k/test_bench/data_lab.json" @@ -0,0 +1,7650 @@ +[ + { + "id": "3006890", + "images": [ + "3006890_00.jpg", + "3006890_01.jpg", + "3006890_02.jpg", + "3006890_03.jpg", + "3006890_04.jpg", + "3006890_05.jpg", + "3006890_06.jpg", + "3006890_07.jpg", + "3006890_08.jpg", + "3006890_09.jpg", + "3006890_10.jpg", + "3006890_11.jpg", + "3006890_12.jpg", + "3006890_13.jpg", + "3006890_14.jpg", + "3006890_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a computer laptop sitting on a beach at sunrise with a volcano erupting from it \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "1006633", + "images": [ + "1006633_00.jpg", + "1006633_01.jpg", + "1006633_02.jpg", + "1006633_03.jpg", + "1006633_04.jpg", + "1006633_05.jpg", + "1006633_06.jpg", + "1006633_07.jpg", + "1006633_08.jpg", + "1006633_09.jpg", + "1006633_10.jpg", + "1006633_11.jpg", + "1006633_12.jpg", + "1006633_13.jpg", + "1006633_14.jpg", + "1006633_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"guy out of a Disney movie with a straw hat \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002504", + "images": [ + "7002504_00.jpg", + "7002504_01.jpg", + "7002504_02.jpg", + "7002504_03.jpg", + "7002504_04.jpg", + "7002504_05.jpg", + "7002504_06.jpg", + "7002504_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"shepherd boy realizing his mistake in forest \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6002691", + "images": [ + "6002691_00.jpg", + "6002691_01.jpg", + "6002691_02.jpg", + "6002691_03.jpg", + "6002691_04.jpg", + "6002691_05.jpg", + "6002691_06.jpg", + "6002691_07.jpg", + "6002691_08.jpg", + "6002691_09.jpg", + "6002691_10.jpg", + "6002691_11.jpg", + "6002691_12.jpg", + "6002691_13.jpg", + "6002691_14.jpg", + "6002691_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A muscular warrior with a taste of a lion running in a desert \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "7003706", + "images": [ + "7003706_00.jpg", + "7003706_01.jpg", + "7003706_02.jpg", + "7003706_03.jpg", + "7003706_04.jpg", + "7003706_05.jpg", + "7003706_06.jpg", + "7003706_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"lord rama returning to ayodhya ancient city animated lighting \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "4004274", + "images": [ + "4004274_00.jpg", + "4004274_01.jpg", + "4004274_02.jpg", + "4004274_03.jpg", + "4004274_04.jpg", + "4004274_05.jpg", + "4004274_06.jpg", + "4004274_07.jpg", + "4004274_08.jpg", + "4004274_09.jpg", + "4004274_10.jpg", + "4004274_11.jpg", + "4004274_12.jpg", + "4004274_13.jpg", + "4004274_14.jpg", + "4004274_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"3d Cartoons Kids dancing in a circle on a cartoony sandy beach, the waves crashing gently in the background, while seagulls fly overhead and the sun sets in a fiery orange sky. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "8002463", + "images": [ + "8002463_00.jpg", + "8002463_01.jpg", + "8002463_02.jpg", + "8002463_03.jpg", + "8002463_04.jpg", + "8002463_05.jpg", + "8002463_06.jpg", + "8002463_07.jpg", + "8002463_08.jpg", + "8002463_09.jpg", + "8002463_10.jpg", + "8002463_11.jpg", + "8002463_12.jpg", + "8002463_13.jpg", + "8002463_14.jpg", + "8002463_15.jpg", + "8002463_16.jpg", + "8002463_17.jpg", + "8002463_18.jpg", + "8002463_19.jpg", + "8002463_20.jpg", + "8002463_21.jpg", + "8002463_22.jpg", + "8002463_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"As the sun began to set, the children returned to the magic playground, their hearts brimming with stories of their incredible journeys \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7003941", + "images": [ + "7003941_00.jpg", + "7003941_01.jpg", + "7003941_02.jpg", + "7003941_03.jpg", + "7003941_04.jpg", + "7003941_05.jpg", + "7003941_06.jpg", + "7003941_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"surrounded by an assortment of pets and animals, including dogs, cats, chickens, and pigeons. camera zoom in,cinematic video, \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "2005219", + "images": [ + "2005219_00.jpg", + "2005219_01.jpg", + "2005219_02.jpg", + "2005219_03.jpg", + "2005219_04.jpg", + "2005219_05.jpg", + "2005219_06.jpg", + "2005219_07.jpg", + "2005219_08.jpg", + "2005219_09.jpg", + "2005219_10.jpg", + "2005219_11.jpg", + "2005219_12.jpg", + "2005219_13.jpg", + "2005219_14.jpg", + "2005219_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a little mermaid in sea with fishes moving happily well dressed clear face \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "2004542", + "images": [ + "2004542_00.jpg", + "2004542_01.jpg", + "2004542_02.jpg", + "2004542_03.jpg", + "2004542_04.jpg", + "2004542_05.jpg", + "2004542_06.jpg", + "2004542_07.jpg", + "2004542_08.jpg", + "2004542_09.jpg", + "2004542_10.jpg", + "2004542_11.jpg", + "2004542_12.jpg", + "2004542_13.jpg", + "2004542_14.jpg", + "2004542_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a cat watching the sea while jumping over rocks \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003539", + "images": [ + "6003539_00.jpg", + "6003539_01.jpg", + "6003539_02.jpg", + "6003539_03.jpg", + "6003539_04.jpg", + "6003539_05.jpg", + "6003539_06.jpg", + "6003539_07.jpg", + "6003539_08.jpg", + "6003539_09.jpg", + "6003539_10.jpg", + "6003539_11.jpg", + "6003539_12.jpg", + "6003539_13.jpg", + "6003539_14.jpg", + "6003539_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"dence forest, mountains, enchanting forest, unreal render, disney style \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004768", + "images": [ + "0004768_00.jpg", + "0004768_01.jpg", + "0004768_02.jpg", + "0004768_03.jpg", + "0004768_04.jpg", + "0004768_05.jpg", + "0004768_06.jpg", + "0004768_07.jpg", + "0004768_08.jpg", + "0004768_09.jpg", + "0004768_10.jpg", + "0004768_11.jpg", + "0004768_12.jpg", + "0004768_13.jpg", + "0004768_14.jpg", + "0004768_15.jpg", + "0004768_16.jpg", + "0004768_17.jpg", + "0004768_18.jpg", + "0004768_19.jpg", + "0004768_20.jpg", + "0004768_21.jpg", + "0004768_22.jpg", + "0004768_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A chocolate brick smashing into a chocolate bar. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "6002546", + "images": [ + "6002546_00.jpg", + "6002546_01.jpg", + "6002546_02.jpg", + "6002546_03.jpg", + "6002546_04.jpg", + "6002546_05.jpg", + "6002546_06.jpg", + "6002546_07.jpg", + "6002546_08.jpg", + "6002546_09.jpg", + "6002546_10.jpg", + "6002546_11.jpg", + "6002546_12.jpg", + "6002546_13.jpg", + "6002546_14.jpg", + "6002546_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Closeup, clearly visible goose down snow falling slowly, blue sky and white clouds background, blizzard, surreal, 3:4 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "0003021", + "images": [ + "0003021_00.jpg", + "0003021_01.jpg", + "0003021_02.jpg", + "0003021_03.jpg", + "0003021_04.jpg", + "0003021_05.jpg", + "0003021_06.jpg", + "0003021_07.jpg", + "0003021_08.jpg", + "0003021_09.jpg", + "0003021_10.jpg", + "0003021_11.jpg", + "0003021_12.jpg", + "0003021_13.jpg", + "0003021_14.jpg", + "0003021_15.jpg", + "0003021_16.jpg", + "0003021_17.jpg", + "0003021_18.jpg", + "0003021_19.jpg", + "0003021_20.jpg", + "0003021_21.jpg", + "0003021_22.jpg", + "0003021_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"old footage from soldier attacking a wolf, cinematic \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3006400", + "images": [ + "3006400_00.jpg", + "3006400_01.jpg", + "3006400_02.jpg", + "3006400_03.jpg", + "3006400_04.jpg", + "3006400_05.jpg", + "3006400_06.jpg", + "3006400_07.jpg", + "3006400_08.jpg", + "3006400_09.jpg", + "3006400_10.jpg", + "3006400_11.jpg", + "3006400_12.jpg", + "3006400_13.jpg", + "3006400_14.jpg", + "3006400_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"portrait of hellraiser cenobite in the style of wayne barlowe, gustav moreau, peter mohrbacher, bussiere, roberto ferri, santiago caruso, luis ricardo falero, austin osman spare \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "2006657", + "images": [ + "2006657_00.jpg", + "2006657_01.jpg", + "2006657_02.jpg", + "2006657_03.jpg", + "2006657_04.jpg", + "2006657_05.jpg", + "2006657_06.jpg", + "2006657_07.jpg", + "2006657_08.jpg", + "2006657_09.jpg", + "2006657_10.jpg", + "2006657_11.jpg", + "2006657_12.jpg", + "2006657_13.jpg", + "2006657_14.jpg", + "2006657_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Gandalalaf He stands in front of the Valley of the Kings tomb wearing the clothes of the Pharaohs, focusing on the facial features and hand movement, and the image is of high quality, cinematic \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0006077", + "images": [ + "0006077_00.jpg", + "0006077_01.jpg", + "0006077_02.jpg", + "0006077_03.jpg", + "0006077_04.jpg", + "0006077_05.jpg", + "0006077_06.jpg", + "0006077_07.jpg", + "0006077_08.jpg", + "0006077_09.jpg", + "0006077_10.jpg", + "0006077_11.jpg", + "0006077_12.jpg", + "0006077_13.jpg", + "0006077_14.jpg", + "0006077_15.jpg", + "0006077_16.jpg", + "0006077_17.jpg", + "0006077_18.jpg", + "0006077_19.jpg", + "0006077_20.jpg", + "0006077_21.jpg", + "0006077_22.jpg", + "0006077_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Dark psychedelic images, in the style of Salvador Dali. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "8003140", + "images": [ + "8003140_00.jpg", + "8003140_01.jpg", + "8003140_02.jpg", + "8003140_03.jpg", + "8003140_04.jpg", + "8003140_05.jpg", + "8003140_06.jpg", + "8003140_07.jpg", + "8003140_08.jpg", + "8003140_09.jpg", + "8003140_10.jpg", + "8003140_11.jpg", + "8003140_12.jpg", + "8003140_13.jpg", + "8003140_14.jpg", + "8003140_15.jpg", + "8003140_16.jpg", + "8003140_17.jpg", + "8003140_18.jpg", + "8003140_19.jpg", + "8003140_20.jpg", + "8003140_21.jpg", + "8003140_22.jpg", + "8003140_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A beautiful girl in the sky on a hippopotamus \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3005993", + "images": [ + "3005993_00.jpg", + "3005993_01.jpg", + "3005993_02.jpg", + "3005993_03.jpg", + "3005993_04.jpg", + "3005993_05.jpg", + "3005993_06.jpg", + "3005993_07.jpg", + "3005993_08.jpg", + "3005993_09.jpg", + "3005993_10.jpg", + "3005993_11.jpg", + "3005993_12.jpg", + "3005993_13.jpg", + "3005993_14.jpg", + "3005993_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Visualize an image of children growing up in space colonies, looking up at the stars and dreaming of new cosmic adventures. It reflects the hope and courage of the new generation. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "7004030", + "images": [ + "7004030_00.jpg", + "7004030_01.jpg", + "7004030_02.jpg", + "7004030_03.jpg", + "7004030_04.jpg", + "7004030_05.jpg", + "7004030_06.jpg", + "7004030_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a viking brian kneeling at prayer \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "2005823", + "images": [ + "2005823_00.jpg", + "2005823_01.jpg", + "2005823_02.jpg", + "2005823_03.jpg", + "2005823_04.jpg", + "2005823_05.jpg", + "2005823_06.jpg", + "2005823_07.jpg", + "2005823_08.jpg", + "2005823_09.jpg", + "2005823_10.jpg", + "2005823_11.jpg", + "2005823_12.jpg", + "2005823_13.jpg", + "2005823_14.jpg", + "2005823_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cinematic ink sketch cartoon of Pepe The Frog slapping a woman with an asymmetrical bob haircut standing in his way, in the style of HR Giger, wide shot, full animation, accurate and detailed, \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "4003452", + "images": [ + "4003452_00.jpg", + "4003452_01.jpg", + "4003452_02.jpg", + "4003452_03.jpg", + "4003452_04.jpg", + "4003452_05.jpg", + "4003452_06.jpg", + "4003452_07.jpg", + "4003452_08.jpg", + "4003452_09.jpg", + "4003452_10.jpg", + "4003452_11.jpg", + "4003452_12.jpg", + "4003452_13.jpg", + "4003452_14.jpg", + "4003452_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"about the mystique of the cosmos, highlighting its profound significance and comparing celestial bodies to beings like ourselves \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "8003483", + "images": [ + "8003483_00.jpg", + "8003483_01.jpg", + "8003483_02.jpg", + "8003483_03.jpg", + "8003483_04.jpg", + "8003483_05.jpg", + "8003483_06.jpg", + "8003483_07.jpg", + "8003483_08.jpg", + "8003483_09.jpg", + "8003483_10.jpg", + "8003483_11.jpg", + "8003483_12.jpg", + "8003483_13.jpg", + "8003483_14.jpg", + "8003483_15.jpg", + "8003483_16.jpg", + "8003483_17.jpg", + "8003483_18.jpg", + "8003483_19.jpg", + "8003483_20.jpg", + "8003483_21.jpg", + "8003483_22.jpg", + "8003483_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"an identity theft scene on an animated web page \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003547", + "images": [ + "6003547_00.jpg", + "6003547_01.jpg", + "6003547_02.jpg", + "6003547_03.jpg", + "6003547_04.jpg", + "6003547_05.jpg", + "6003547_06.jpg", + "6003547_07.jpg", + "6003547_08.jpg", + "6003547_09.jpg", + "6003547_10.jpg", + "6003547_11.jpg", + "6003547_12.jpg", + "6003547_13.jpg", + "6003547_14.jpg", + "6003547_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a horse galopping through the mainstreet of a western town, wide shot \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2005053", + "images": [ + "2005053_00.jpg", + "2005053_01.jpg", + "2005053_02.jpg", + "2005053_03.jpg", + "2005053_04.jpg", + "2005053_05.jpg", + "2005053_06.jpg", + "2005053_07.jpg", + "2005053_08.jpg", + "2005053_09.jpg", + "2005053_10.jpg", + "2005053_11.jpg", + "2005053_12.jpg", + "2005053_13.jpg", + "2005053_14.jpg", + "2005053_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"anime teenager with red hair in classrom and he wearing black costume \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004906", + "images": [ + "0004906_00.jpg", + "0004906_01.jpg", + "0004906_02.jpg", + "0004906_03.jpg", + "0004906_04.jpg", + "0004906_05.jpg", + "0004906_06.jpg", + "0004906_07.jpg", + "0004906_08.jpg", + "0004906_09.jpg", + "0004906_10.jpg", + "0004906_11.jpg", + "0004906_12.jpg", + "0004906_13.jpg", + "0004906_14.jpg", + "0004906_15.jpg", + "0004906_16.jpg", + "0004906_17.jpg", + "0004906_18.jpg", + "0004906_19.jpg", + "0004906_20.jpg", + "0004906_21.jpg", + "0004906_22.jpg", + "0004906_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A healthcare worker tending to gas victims \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "3004512", + "images": [ + "3004512_00.jpg", + "3004512_01.jpg", + "3004512_02.jpg", + "3004512_03.jpg", + "3004512_04.jpg", + "3004512_05.jpg", + "3004512_06.jpg", + "3004512_07.jpg", + "3004512_08.jpg", + "3004512_09.jpg", + "3004512_10.jpg", + "3004512_11.jpg", + "3004512_12.jpg", + "3004512_13.jpg", + "3004512_14.jpg", + "3004512_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a man in a suit walks away from a private jet, a digital rendering, by Robert Gavin, gold and blue color scheme, side view of a gaunt, new vehicle \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "4004914", + "images": [ + "4004914_00.jpg", + "4004914_01.jpg", + "4004914_02.jpg", + "4004914_03.jpg", + "4004914_04.jpg", + "4004914_05.jpg", + "4004914_06.jpg", + "4004914_07.jpg", + "4004914_08.jpg", + "4004914_09.jpg", + "4004914_10.jpg", + "4004914_11.jpg", + "4004914_12.jpg", + "4004914_13.jpg", + "4004914_14.jpg", + "4004914_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Silhouetted against a moonlit sky, ninjas leap between rooftops, their blades flashing, while cherry blossom petals whirl in their slipstream. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "2004947", + "images": [ + "2004947_00.jpg", + "2004947_01.jpg", + "2004947_02.jpg", + "2004947_03.jpg", + "2004947_04.jpg", + "2004947_05.jpg", + "2004947_06.jpg", + "2004947_07.jpg", + "2004947_08.jpg", + "2004947_09.jpg", + "2004947_10.jpg", + "2004947_11.jpg", + "2004947_12.jpg", + "2004947_13.jpg", + "2004947_14.jpg", + "2004947_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A girl holding a green tablet \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "7004580", + "images": [ + "7004580_00.jpg", + "7004580_01.jpg", + "7004580_02.jpg", + "7004580_03.jpg", + "7004580_04.jpg", + "7004580_05.jpg", + "7004580_06.jpg", + "7004580_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a man riding a bike on the road, look very happy, backround is taipei city \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "6003985", + "images": [ + "6003985_00.jpg", + "6003985_01.jpg", + "6003985_02.jpg", + "6003985_03.jpg", + "6003985_04.jpg", + "6003985_05.jpg", + "6003985_06.jpg", + "6003985_07.jpg", + "6003985_08.jpg", + "6003985_09.jpg", + "6003985_10.jpg", + "6003985_11.jpg", + "6003985_12.jpg", + "6003985_13.jpg", + "6003985_14.jpg", + "6003985_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Description: Show a traditional Aarti (ritual) being performed after tying the Rakhi, with a glowing diya (lamp) and incense.\nSymbolism: Highlight the sacredness of the Rakhi ceremony and the blessings being bestowed.\nBackground: A serene, softly lit environment to convey the spirituality of the moment.\nCamera Position: A focused view of the Aarti with the Rakhi in the background.\nVideo Quality: Create a tranquil and spiritually uplifting atmosphere. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "8003500", + "images": [ + "8003500_00.jpg", + "8003500_01.jpg", + "8003500_02.jpg", + "8003500_03.jpg", + "8003500_04.jpg", + "8003500_05.jpg", + "8003500_06.jpg", + "8003500_07.jpg", + "8003500_08.jpg", + "8003500_09.jpg", + "8003500_10.jpg", + "8003500_11.jpg", + "8003500_12.jpg", + "8003500_13.jpg", + "8003500_14.jpg", + "8003500_15.jpg", + "8003500_16.jpg", + "8003500_17.jpg", + "8003500_18.jpg", + "8003500_19.jpg", + "8003500_20.jpg", + "8003500_21.jpg", + "8003500_22.jpg", + "8003500_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"boy hard work and dedication made him successful in the All India Medical Entrance Examination. He became a prominent doctor \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004229", + "images": [ + "0004229_00.jpg", + "0004229_01.jpg", + "0004229_02.jpg", + "0004229_03.jpg", + "0004229_04.jpg", + "0004229_05.jpg", + "0004229_06.jpg", + "0004229_07.jpg", + "0004229_08.jpg", + "0004229_09.jpg", + "0004229_10.jpg", + "0004229_11.jpg", + "0004229_12.jpg", + "0004229_13.jpg", + "0004229_14.jpg", + "0004229_15.jpg", + "0004229_16.jpg", + "0004229_17.jpg", + "0004229_18.jpg", + "0004229_19.jpg", + "0004229_20.jpg", + "0004229_21.jpg", + "0004229_22.jpg", + "0004229_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a pack of wolves running through a forest at night with a huge full moon in the sky. Exaggerated wide angle, high contrast, 16:9 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "1004206", + "images": [ + "1004206_00.jpg", + "1004206_01.jpg", + "1004206_02.jpg", + "1004206_03.jpg", + "1004206_04.jpg", + "1004206_05.jpg", + "1004206_06.jpg", + "1004206_07.jpg", + "1004206_08.jpg", + "1004206_09.jpg", + "1004206_10.jpg", + "1004206_11.jpg", + "1004206_12.jpg", + "1004206_13.jpg", + "1004206_14.jpg", + "1004206_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The fox approached the base of the tree and looked up, his eyes fixed on the crow \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3005976", + "images": [ + "3005976_00.jpg", + "3005976_01.jpg", + "3005976_02.jpg", + "3005976_03.jpg", + "3005976_04.jpg", + "3005976_05.jpg", + "3005976_06.jpg", + "3005976_07.jpg", + "3005976_08.jpg", + "3005976_09.jpg", + "3005976_10.jpg", + "3005976_11.jpg", + "3005976_12.jpg", + "3005976_13.jpg", + "3005976_14.jpg", + "3005976_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cinematic,very detailed,people dancing in the background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004072", + "images": [ + "0004072_00.jpg", + "0004072_01.jpg", + "0004072_02.jpg", + "0004072_03.jpg", + "0004072_04.jpg", + "0004072_05.jpg", + "0004072_06.jpg", + "0004072_07.jpg", + "0004072_08.jpg", + "0004072_09.jpg", + "0004072_10.jpg", + "0004072_11.jpg", + "0004072_12.jpg", + "0004072_13.jpg", + "0004072_14.jpg", + "0004072_15.jpg", + "0004072_16.jpg", + "0004072_17.jpg", + "0004072_18.jpg", + "0004072_19.jpg", + "0004072_20.jpg", + "0004072_21.jpg", + "0004072_22.jpg", + "0004072_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"wide angle from the balcony: looking the landscape with Giger palaces in ruins, satanic monuments, totems of evil creatures, dramatic cloudy sky, mystical, hyperdetailed, intricate perfect details, complex, horror art. Illustration by Dan Seagrave \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "2003121", + "images": [ + "2003121_00.jpg", + "2003121_01.jpg", + "2003121_02.jpg", + "2003121_03.jpg", + "2003121_04.jpg", + "2003121_05.jpg", + "2003121_06.jpg", + "2003121_07.jpg", + "2003121_08.jpg", + "2003121_09.jpg", + "2003121_10.jpg", + "2003121_11.jpg", + "2003121_12.jpg", + "2003121_13.jpg", + "2003121_14.jpg", + "2003121_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"old liminal russian cemetery with two ghosts dancing with each other 1950\u2019s black and white film \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1004245", + "images": [ + "1004245_00.jpg", + "1004245_01.jpg", + "1004245_02.jpg", + "1004245_03.jpg", + "1004245_04.jpg", + "1004245_05.jpg", + "1004245_06.jpg", + "1004245_07.jpg", + "1004245_08.jpg", + "1004245_09.jpg", + "1004245_10.jpg", + "1004245_11.jpg", + "1004245_12.jpg", + "1004245_13.jpg", + "1004245_14.jpg", + "1004245_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"staying hydrated during exercise can enhance performance and calorieburn. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004307", + "images": [ + "0004307_00.jpg", + "0004307_01.jpg", + "0004307_02.jpg", + "0004307_03.jpg", + "0004307_04.jpg", + "0004307_05.jpg", + "0004307_06.jpg", + "0004307_07.jpg", + "0004307_08.jpg", + "0004307_09.jpg", + "0004307_10.jpg", + "0004307_11.jpg", + "0004307_12.jpg", + "0004307_13.jpg", + "0004307_14.jpg", + "0004307_15.jpg", + "0004307_16.jpg", + "0004307_17.jpg", + "0004307_18.jpg", + "0004307_19.jpg", + "0004307_20.jpg", + "0004307_21.jpg", + "0004307_22.jpg", + "0004307_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"3D animated video of john and meeshu travelling in a bicycle inside a deep forest \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3006681", + "images": [ + "3006681_00.jpg", + "3006681_01.jpg", + "3006681_02.jpg", + "3006681_03.jpg", + "3006681_04.jpg", + "3006681_05.jpg", + "3006681_06.jpg", + "3006681_07.jpg", + "3006681_08.jpg", + "3006681_09.jpg", + "3006681_10.jpg", + "3006681_11.jpg", + "3006681_12.jpg", + "3006681_13.jpg", + "3006681_14.jpg", + "3006681_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Jennifer Lopez mowing the lawn wearing a Washington Redskins bikini. Anime style. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1003674", + "images": [ + "1003674_00.jpg", + "1003674_01.jpg", + "1003674_02.jpg", + "1003674_03.jpg", + "1003674_04.jpg", + "1003674_05.jpg", + "1003674_06.jpg", + "1003674_07.jpg", + "1003674_08.jpg", + "1003674_09.jpg", + "1003674_10.jpg", + "1003674_11.jpg", + "1003674_12.jpg", + "1003674_13.jpg", + "1003674_14.jpg", + "1003674_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Joe Biden, red lazer eyes, surging with power \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003510", + "images": [ + "6003510_00.jpg", + "6003510_01.jpg", + "6003510_02.jpg", + "6003510_03.jpg", + "6003510_04.jpg", + "6003510_05.jpg", + "6003510_06.jpg", + "6003510_07.jpg", + "6003510_08.jpg", + "6003510_09.jpg", + "6003510_10.jpg", + "6003510_11.jpg", + "6003510_12.jpg", + "6003510_13.jpg", + "6003510_14.jpg", + "6003510_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"barack obama vigerusly dancing in the rural woods 8k realistic \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "2004327", + "images": [ + "2004327_00.jpg", + "2004327_01.jpg", + "2004327_02.jpg", + "2004327_03.jpg", + "2004327_04.jpg", + "2004327_05.jpg", + "2004327_06.jpg", + "2004327_07.jpg", + "2004327_08.jpg", + "2004327_09.jpg", + "2004327_10.jpg", + "2004327_11.jpg", + "2004327_12.jpg", + "2004327_13.jpg", + "2004327_14.jpg", + "2004327_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"super earthquake with a lot of building collapsing \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "0003433", + "images": [ + "0003433_00.jpg", + "0003433_01.jpg", + "0003433_02.jpg", + "0003433_03.jpg", + "0003433_04.jpg", + "0003433_05.jpg", + "0003433_06.jpg", + "0003433_07.jpg", + "0003433_08.jpg", + "0003433_09.jpg", + "0003433_10.jpg", + "0003433_11.jpg", + "0003433_12.jpg", + "0003433_13.jpg", + "0003433_14.jpg", + "0003433_15.jpg", + "0003433_16.jpg", + "0003433_17.jpg", + "0003433_18.jpg", + "0003433_19.jpg", + "0003433_20.jpg", + "0003433_21.jpg", + "0003433_22.jpg", + "0003433_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a video showing rocks falling off the side of mountain onto a road curving alongside the mountain \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002407", + "images": [ + "7002407_00.jpg", + "7002407_01.jpg", + "7002407_02.jpg", + "7002407_03.jpg", + "7002407_04.jpg", + "7002407_05.jpg", + "7002407_06.jpg", + "7002407_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Man dressed as mario eating spaghettimeatballs at a wedding \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1005303", + "images": [ + "1005303_00.jpg", + "1005303_01.jpg", + "1005303_02.jpg", + "1005303_03.jpg", + "1005303_04.jpg", + "1005303_05.jpg", + "1005303_06.jpg", + "1005303_07.jpg", + "1005303_08.jpg", + "1005303_09.jpg", + "1005303_10.jpg", + "1005303_11.jpg", + "1005303_12.jpg", + "1005303_13.jpg", + "1005303_14.jpg", + "1005303_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A dinosaur walking towards a modern villa, low angle shot \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3004776", + "images": [ + "3004776_00.jpg", + "3004776_01.jpg", + "3004776_02.jpg", + "3004776_03.jpg", + "3004776_04.jpg", + "3004776_05.jpg", + "3004776_06.jpg", + "3004776_07.jpg", + "3004776_08.jpg", + "3004776_09.jpg", + "3004776_10.jpg", + "3004776_11.jpg", + "3004776_12.jpg", + "3004776_13.jpg", + "3004776_14.jpg", + "3004776_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"night, glitch art, background only, analogue errors, pixels, corrupted digital data \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1005423", + "images": [ + "1005423_00.jpg", + "1005423_01.jpg", + "1005423_02.jpg", + "1005423_03.jpg", + "1005423_04.jpg", + "1005423_05.jpg", + "1005423_06.jpg", + "1005423_07.jpg", + "1005423_08.jpg", + "1005423_09.jpg", + "1005423_10.jpg", + "1005423_11.jpg", + "1005423_12.jpg", + "1005423_13.jpg", + "1005423_14.jpg", + "1005423_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"kid praying in mosque with motion \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2004932", + "images": [ + "2004932_00.jpg", + "2004932_01.jpg", + "2004932_02.jpg", + "2004932_03.jpg", + "2004932_04.jpg", + "2004932_05.jpg", + "2004932_06.jpg", + "2004932_07.jpg", + "2004932_08.jpg", + "2004932_09.jpg", + "2004932_10.jpg", + "2004932_11.jpg", + "2004932_12.jpg", + "2004932_13.jpg", + "2004932_14.jpg", + "2004932_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a man being resured from a deep dark pit \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2006473", + "images": [ + "2006473_00.jpg", + "2006473_01.jpg", + "2006473_02.jpg", + "2006473_03.jpg", + "2006473_04.jpg", + "2006473_05.jpg", + "2006473_06.jpg", + "2006473_07.jpg", + "2006473_08.jpg", + "2006473_09.jpg", + "2006473_10.jpg", + "2006473_11.jpg", + "2006473_12.jpg", + "2006473_13.jpg", + "2006473_14.jpg", + "2006473_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"create a 5 min football video \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3006692", + "images": [ + "3006692_00.jpg", + "3006692_01.jpg", + "3006692_02.jpg", + "3006692_03.jpg", + "3006692_04.jpg", + "3006692_05.jpg", + "3006692_06.jpg", + "3006692_07.jpg", + "3006692_08.jpg", + "3006692_09.jpg", + "3006692_10.jpg", + "3006692_11.jpg", + "3006692_12.jpg", + "3006692_13.jpg", + "3006692_14.jpg", + "3006692_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a happy kid jumping on the park \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001551", + "images": [ + "8001551_00.jpg", + "8001551_01.jpg", + "8001551_02.jpg", + "8001551_03.jpg", + "8001551_04.jpg", + "8001551_05.jpg", + "8001551_06.jpg", + "8001551_07.jpg", + "8001551_08.jpg", + "8001551_09.jpg", + "8001551_10.jpg", + "8001551_11.jpg", + "8001551_12.jpg", + "8001551_13.jpg", + "8001551_14.jpg", + "8001551_15.jpg", + "8001551_16.jpg", + "8001551_17.jpg", + "8001551_18.jpg", + "8001551_19.jpg", + "8001551_20.jpg", + "8001551_21.jpg", + "8001551_22.jpg", + "8001551_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"13 glasses with different colored drinks in a bar with neon lights Modern futuristic comic style \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "8003849", + "images": [ + "8003849_00.jpg", + "8003849_01.jpg", + "8003849_02.jpg", + "8003849_03.jpg", + "8003849_04.jpg", + "8003849_05.jpg", + "8003849_06.jpg", + "8003849_07.jpg", + "8003849_08.jpg", + "8003849_09.jpg", + "8003849_10.jpg", + "8003849_11.jpg", + "8003849_12.jpg", + "8003849_13.jpg", + "8003849_14.jpg", + "8003849_15.jpg", + "8003849_16.jpg", + "8003849_17.jpg", + "8003849_18.jpg", + "8003849_19.jpg", + "8003849_20.jpg", + "8003849_21.jpg", + "8003849_22.jpg", + "8003849_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The guy is sitting in the background of an abandoned city and the camera moves away \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "7003885", + "images": [ + "7003885_00.jpg", + "7003885_01.jpg", + "7003885_02.jpg", + "7003885_03.jpg", + "7003885_04.jpg", + "7003885_05.jpg", + "7003885_06.jpg", + "7003885_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"\",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "6002885", + "images": [ + "6002885_00.jpg", + "6002885_01.jpg", + "6002885_02.jpg", + "6002885_03.jpg", + "6002885_04.jpg", + "6002885_05.jpg", + "6002885_06.jpg", + "6002885_07.jpg", + "6002885_08.jpg", + "6002885_09.jpg", + "6002885_10.jpg", + "6002885_11.jpg", + "6002885_12.jpg", + "6002885_13.jpg", + "6002885_14.jpg", + "6002885_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a farmer who lost his watch \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0003884", + "images": [ + "0003884_00.jpg", + "0003884_01.jpg", + "0003884_02.jpg", + "0003884_03.jpg", + "0003884_04.jpg", + "0003884_05.jpg", + "0003884_06.jpg", + "0003884_07.jpg", + "0003884_08.jpg", + "0003884_09.jpg", + "0003884_10.jpg", + "0003884_11.jpg", + "0003884_12.jpg", + "0003884_13.jpg", + "0003884_14.jpg", + "0003884_15.jpg", + "0003884_16.jpg", + "0003884_17.jpg", + "0003884_18.jpg", + "0003884_19.jpg", + "0003884_20.jpg", + "0003884_21.jpg", + "0003884_22.jpg", + "0003884_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"fast flying forward through the galaxy \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "2003144", + "images": [ + "2003144_00.jpg", + "2003144_01.jpg", + "2003144_02.jpg", + "2003144_03.jpg", + "2003144_04.jpg", + "2003144_05.jpg", + "2003144_06.jpg", + "2003144_07.jpg", + "2003144_08.jpg", + "2003144_09.jpg", + "2003144_10.jpg", + "2003144_11.jpg", + "2003144_12.jpg", + "2003144_13.jpg", + "2003144_14.jpg", + "2003144_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a cat stood tall with the body of a human, but the face of a cat, standing tall \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0006762", + "images": [ + "0006762_00.jpg", + "0006762_01.jpg", + "0006762_02.jpg", + "0006762_03.jpg", + "0006762_04.jpg", + "0006762_05.jpg", + "0006762_06.jpg", + "0006762_07.jpg", + "0006762_08.jpg", + "0006762_09.jpg", + "0006762_10.jpg", + "0006762_11.jpg", + "0006762_12.jpg", + "0006762_13.jpg", + "0006762_14.jpg", + "0006762_15.jpg", + "0006762_16.jpg", + "0006762_17.jpg", + "0006762_18.jpg", + "0006762_19.jpg", + "0006762_20.jpg", + "0006762_21.jpg", + "0006762_22.jpg", + "0006762_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"saturn rings around the earth zoom out video \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "1003702", + "images": [ + "1003702_00.jpg", + "1003702_01.jpg", + "1003702_02.jpg", + "1003702_03.jpg", + "1003702_04.jpg", + "1003702_05.jpg", + "1003702_06.jpg", + "1003702_07.jpg", + "1003702_08.jpg", + "1003702_09.jpg", + "1003702_10.jpg", + "1003702_11.jpg", + "1003702_12.jpg", + "1003702_13.jpg", + "1003702_14.jpg", + "1003702_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a lone astronaut crossing the desert at sunset. wide angle overhead shot. 4k. golden hour. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7002951", + "images": [ + "7002951_00.jpg", + "7002951_01.jpg", + "7002951_02.jpg", + "7002951_03.jpg", + "7002951_04.jpg", + "7002951_05.jpg", + "7002951_06.jpg", + "7002951_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"entered the other room, he saw that it was dark and had only a table and a few chairs, camera Pan zoom \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "7004987", + "images": [ + "7004987_00.jpg", + "7004987_01.jpg", + "7004987_02.jpg", + "7004987_03.jpg", + "7004987_04.jpg", + "7004987_05.jpg", + "7004987_06.jpg", + "7004987_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"\",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "4003740", + "images": [ + "4003740_00.jpg", + "4003740_01.jpg", + "4003740_02.jpg", + "4003740_03.jpg", + "4003740_04.jpg", + "4003740_05.jpg", + "4003740_06.jpg", + "4003740_07.jpg", + "4003740_08.jpg", + "4003740_09.jpg", + "4003740_10.jpg", + "4003740_11.jpg", + "4003740_12.jpg", + "4003740_13.jpg", + "4003740_14.jpg", + "4003740_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"light bulbs glowing on n off \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7003310", + "images": [ + "7003310_00.jpg", + "7003310_01.jpg", + "7003310_02.jpg", + "7003310_03.jpg", + "7003310_04.jpg", + "7003310_05.jpg", + "7003310_06.jpg", + "7003310_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"golden skeletons rapping into a fisheye camera \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "5001691", + "images": [ + "5001691_00.jpg", + "5001691_01.jpg", + "5001691_02.jpg", + "5001691_03.jpg", + "5001691_04.jpg", + "5001691_05.jpg", + "5001691_06.jpg", + "5001691_07.jpg", + "5001691_08.jpg", + "5001691_09.jpg", + "5001691_10.jpg", + "5001691_11.jpg", + "5001691_12.jpg", + "5001691_13.jpg", + "5001691_14.jpg", + "5001691_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Mysteriously traveling through the tunnel, the surrounding areas of the tunnel are moving forward rapidly \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "4002006", + "images": [ + "4002006_00.jpg", + "4002006_01.jpg", + "4002006_02.jpg", + "4002006_03.jpg", + "4002006_04.jpg", + "4002006_05.jpg", + "4002006_06.jpg", + "4002006_07.jpg", + "4002006_08.jpg", + "4002006_09.jpg", + "4002006_10.jpg", + "4002006_11.jpg", + "4002006_12.jpg", + "4002006_13.jpg", + "4002006_14.jpg", + "4002006_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Woman in bikini cutting the finish tape on track \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "6004370", + "images": [ + "6004370_00.jpg", + "6004370_01.jpg", + "6004370_02.jpg", + "6004370_03.jpg", + "6004370_04.jpg", + "6004370_05.jpg", + "6004370_06.jpg", + "6004370_07.jpg", + "6004370_08.jpg", + "6004370_09.jpg", + "6004370_10.jpg", + "6004370_11.jpg", + "6004370_12.jpg", + "6004370_13.jpg", + "6004370_14.jpg", + "6004370_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Camera opens to a dimly lit, cluttered bathroom. A young woman, LUCY, stands in front of a large ornate mirror, cleaning it. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6004175", + "images": [ + "6004175_00.jpg", + "6004175_01.jpg", + "6004175_02.jpg", + "6004175_03.jpg", + "6004175_04.jpg", + "6004175_05.jpg", + "6004175_06.jpg", + "6004175_07.jpg", + "6004175_08.jpg", + "6004175_09.jpg", + "6004175_10.jpg", + "6004175_11.jpg", + "6004175_12.jpg", + "6004175_13.jpg", + "6004175_14.jpg", + "6004175_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Edvard Munch animation of a man watching television in nature \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "6004655", + "images": [ + "6004655_00.jpg", + "6004655_01.jpg", + "6004655_02.jpg", + "6004655_03.jpg", + "6004655_04.jpg", + "6004655_05.jpg", + "6004655_06.jpg", + "6004655_07.jpg", + "6004655_08.jpg", + "6004655_09.jpg", + "6004655_10.jpg", + "6004655_11.jpg", + "6004655_12.jpg", + "6004655_13.jpg", + "6004655_14.jpg", + "6004655_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The revolution begins, the peasants rise up against the government, they are strong but unarmed \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "2006775", + "images": [ + "2006775_00.jpg", + "2006775_01.jpg", + "2006775_02.jpg", + "2006775_03.jpg", + "2006775_04.jpg", + "2006775_05.jpg", + "2006775_06.jpg", + "2006775_07.jpg", + "2006775_08.jpg", + "2006775_09.jpg", + "2006775_10.jpg", + "2006775_11.jpg", + "2006775_12.jpg", + "2006775_13.jpg", + "2006775_14.jpg", + "2006775_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"In the bustling world of animated insects, our diligent cartoon ant and carefree grasshopper find themselves preparing for the annual Bug Festival. While the ant diligently collects supplies for the winter, the grasshopper is busy enjoying the festivities. How do their contrasting approaches to life lead to unexpected challenges and heartwarming lessons about the importance of balance and preparation? \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "6004924", + "images": [ + "6004924_00.jpg", + "6004924_01.jpg", + "6004924_02.jpg", + "6004924_03.jpg", + "6004924_04.jpg", + "6004924_05.jpg", + "6004924_06.jpg", + "6004924_07.jpg", + "6004924_08.jpg", + "6004924_09.jpg", + "6004924_10.jpg", + "6004924_11.jpg", + "6004924_12.jpg", + "6004924_13.jpg", + "6004924_14.jpg", + "6004924_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a man fleeing a circus film Lovecraft 1900s \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "0005352", + "images": [ + "0005352_00.jpg", + "0005352_01.jpg", + "0005352_02.jpg", + "0005352_03.jpg", + "0005352_04.jpg", + "0005352_05.jpg", + "0005352_06.jpg", + "0005352_07.jpg", + "0005352_08.jpg", + "0005352_09.jpg", + "0005352_10.jpg", + "0005352_11.jpg", + "0005352_12.jpg", + "0005352_13.jpg", + "0005352_14.jpg", + "0005352_15.jpg", + "0005352_16.jpg", + "0005352_17.jpg", + "0005352_18.jpg", + "0005352_19.jpg", + "0005352_20.jpg", + "0005352_21.jpg", + "0005352_22.jpg", + "0005352_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Sofia engages in intense combat with a horde of sleek, menacing robots. Her movements are swift and calculated as she battles against overwhelming odds. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "7004486", + "images": [ + "7004486_00.jpg", + "7004486_01.jpg", + "7004486_02.jpg", + "7004486_03.jpg", + "7004486_04.jpg", + "7004486_05.jpg", + "7004486_06.jpg", + "7004486_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"3d animation, thick south american jungle. thick mist, large trees and bushes cover the floor. scenery is draped in shadows from the overhead canopy. in the background is a set of menacing eyes peering through a bush \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0004033", + "images": [ + "0004033_00.jpg", + "0004033_01.jpg", + "0004033_02.jpg", + "0004033_03.jpg", + "0004033_04.jpg", + "0004033_05.jpg", + "0004033_06.jpg", + "0004033_07.jpg", + "0004033_08.jpg", + "0004033_09.jpg", + "0004033_10.jpg", + "0004033_11.jpg", + "0004033_12.jpg", + "0004033_13.jpg", + "0004033_14.jpg", + "0004033_15.jpg", + "0004033_16.jpg", + "0004033_17.jpg", + "0004033_18.jpg", + "0004033_19.jpg", + "0004033_20.jpg", + "0004033_21.jpg", + "0004033_22.jpg", + "0004033_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a ghost carrying a mining device in the smoke with Many fluorescent colors \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "3005276", + "images": [ + "3005276_00.jpg", + "3005276_01.jpg", + "3005276_02.jpg", + "3005276_03.jpg", + "3005276_04.jpg", + "3005276_05.jpg", + "3005276_06.jpg", + "3005276_07.jpg", + "3005276_08.jpg", + "3005276_09.jpg", + "3005276_10.jpg", + "3005276_11.jpg", + "3005276_12.jpg", + "3005276_13.jpg", + "3005276_14.jpg", + "3005276_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"skyline city Waking up in the morning \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "1006253", + "images": [ + "1006253_00.jpg", + "1006253_01.jpg", + "1006253_02.jpg", + "1006253_03.jpg", + "1006253_04.jpg", + "1006253_05.jpg", + "1006253_06.jpg", + "1006253_07.jpg", + "1006253_08.jpg", + "1006253_09.jpg", + "1006253_10.jpg", + "1006253_11.jpg", + "1006253_12.jpg", + "1006253_13.jpg", + "1006253_14.jpg", + "1006253_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"eagle sitting on the top of a mountain speaking \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "4002910", + "images": [ + "4002910_00.jpg", + "4002910_01.jpg", + "4002910_02.jpg", + "4002910_03.jpg", + "4002910_04.jpg", + "4002910_05.jpg", + "4002910_06.jpg", + "4002910_07.jpg", + "4002910_08.jpg", + "4002910_09.jpg", + "4002910_10.jpg", + "4002910_11.jpg", + "4002910_12.jpg", + "4002910_13.jpg", + "4002910_14.jpg", + "4002910_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A cute little bird flew to the window sill of a house \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0006269", + "images": [ + "0006269_00.jpg", + "0006269_01.jpg", + "0006269_02.jpg", + "0006269_03.jpg", + "0006269_04.jpg", + "0006269_05.jpg", + "0006269_06.jpg", + "0006269_07.jpg", + "0006269_08.jpg", + "0006269_09.jpg", + "0006269_10.jpg", + "0006269_11.jpg", + "0006269_12.jpg", + "0006269_13.jpg", + "0006269_14.jpg", + "0006269_15.jpg", + "0006269_16.jpg", + "0006269_17.jpg", + "0006269_18.jpg", + "0006269_19.jpg", + "0006269_20.jpg", + "0006269_21.jpg", + "0006269_22.jpg", + "0006269_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"the hero say hi on the 2nd floor of the butiful building \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "0004282", + "images": [ + "0004282_00.jpg", + "0004282_01.jpg", + "0004282_02.jpg", + "0004282_03.jpg", + "0004282_04.jpg", + "0004282_05.jpg", + "0004282_06.jpg", + "0004282_07.jpg", + "0004282_08.jpg", + "0004282_09.jpg", + "0004282_10.jpg", + "0004282_11.jpg", + "0004282_12.jpg", + "0004282_13.jpg", + "0004282_14.jpg", + "0004282_15.jpg", + "0004282_16.jpg", + "0004282_17.jpg", + "0004282_18.jpg", + "0004282_19.jpg", + "0004282_20.jpg", + "0004282_21.jpg", + "0004282_22.jpg", + "0004282_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"steak bun steamy table shot tasty food \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "8002976", + "images": [ + "8002976_00.jpg", + "8002976_01.jpg", + "8002976_02.jpg", + "8002976_03.jpg", + "8002976_04.jpg", + "8002976_05.jpg", + "8002976_06.jpg", + "8002976_07.jpg", + "8002976_08.jpg", + "8002976_09.jpg", + "8002976_10.jpg", + "8002976_11.jpg", + "8002976_12.jpg", + "8002976_13.jpg", + "8002976_14.jpg", + "8002976_15.jpg", + "8002976_16.jpg", + "8002976_17.jpg", + "8002976_18.jpg", + "8002976_19.jpg", + "8002976_20.jpg", + "8002976_21.jpg", + "8002976_22.jpg", + "8002976_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Alice was walking home from school when she saw something shiny on the sidewalk. She picked it up and realized it was a wallet. She opened it and found a driver\u2019s license, a credit card, and some cash. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3003742", + "images": [ + "3003742_00.jpg", + "3003742_01.jpg", + "3003742_02.jpg", + "3003742_03.jpg", + "3003742_04.jpg", + "3003742_05.jpg", + "3003742_06.jpg", + "3003742_07.jpg", + "3003742_08.jpg", + "3003742_09.jpg", + "3003742_10.jpg", + "3003742_11.jpg", + "3003742_12.jpg", + "3003742_13.jpg", + "3003742_14.jpg", + "3003742_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"stunning club party laser show inspired by starwars \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "4002459", + "images": [ + "4002459_00.jpg", + "4002459_01.jpg", + "4002459_02.jpg", + "4002459_03.jpg", + "4002459_04.jpg", + "4002459_05.jpg", + "4002459_06.jpg", + "4002459_07.jpg", + "4002459_08.jpg", + "4002459_09.jpg", + "4002459_10.jpg", + "4002459_11.jpg", + "4002459_12.jpg", + "4002459_13.jpg", + "4002459_14.jpg", + "4002459_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"video of blonde influencer promoting her secret beauty boxes, smiling, showing the pink boxes, numbers on the screen \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "7002232", + "images": [ + "7002232_00.jpg", + "7002232_01.jpg", + "7002232_02.jpg", + "7002232_03.jpg", + "7002232_04.jpg", + "7002232_05.jpg", + "7002232_06.jpg", + "7002232_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"In the dim light of a dark alley, where shadows danced like silent accomplices, emerged the enigmatic Illusionist. Dressed in black robes that seemed to absorb light, his face remained concealed by a dark cloak, only revealing a sinister smile that would suggest the promise of dark secrets. With piercing eyes that sparkled with an unearthly brilliance, he manipulated the imaginations of the unwary young people who crossed his path. Her words were like hypnotic spells, whispered \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "6004514", + "images": [ + "6004514_00.jpg", + "6004514_01.jpg", + "6004514_02.jpg", + "6004514_03.jpg", + "6004514_04.jpg", + "6004514_05.jpg", + "6004514_06.jpg", + "6004514_07.jpg", + "6004514_08.jpg", + "6004514_09.jpg", + "6004514_10.jpg", + "6004514_11.jpg", + "6004514_12.jpg", + "6004514_13.jpg", + "6004514_14.jpg", + "6004514_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Illustrate the shooting star streaking across the night sky as the squirrel makes a wish with a hopeful expression \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0003959", + "images": [ + "0003959_00.jpg", + "0003959_01.jpg", + "0003959_02.jpg", + "0003959_03.jpg", + "0003959_04.jpg", + "0003959_05.jpg", + "0003959_06.jpg", + "0003959_07.jpg", + "0003959_08.jpg", + "0003959_09.jpg", + "0003959_10.jpg", + "0003959_11.jpg", + "0003959_12.jpg", + "0003959_13.jpg", + "0003959_14.jpg", + "0003959_15.jpg", + "0003959_16.jpg", + "0003959_17.jpg", + "0003959_18.jpg", + "0003959_19.jpg", + "0003959_20.jpg", + "0003959_21.jpg", + "0003959_22.jpg", + "0003959_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"full shot of person eating a big juicy hamburger with thick meat, yellow cheese, in the shot that you can see the person sitting, holding the hamburger with both hands while biting it and dripping the sauce enjoying eating it \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0005560", + "images": [ + "0005560_00.jpg", + "0005560_01.jpg", + "0005560_02.jpg", + "0005560_03.jpg", + "0005560_04.jpg", + "0005560_05.jpg", + "0005560_06.jpg", + "0005560_07.jpg", + "0005560_08.jpg", + "0005560_09.jpg", + "0005560_10.jpg", + "0005560_11.jpg", + "0005560_12.jpg", + "0005560_13.jpg", + "0005560_14.jpg", + "0005560_15.jpg", + "0005560_16.jpg", + "0005560_17.jpg", + "0005560_18.jpg", + "0005560_19.jpg", + "0005560_20.jpg", + "0005560_21.jpg", + "0005560_22.jpg", + "0005560_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a woman middle aged, brown hair and warm face she is walking alone at dusk on a wooded trail in fall, she is worried she is being followed \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002872", + "images": [ + "7002872_00.jpg", + "7002872_01.jpg", + "7002872_02.jpg", + "7002872_03.jpg", + "7002872_04.jpg", + "7002872_05.jpg", + "7002872_06.jpg", + "7002872_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Parashurama traveled across the land, challenging and defeating numerous kings and warriors. He became a legendary figure known for his relentless pursuit of justice. He fought countless battles, wielding his mighty axe and displaying extraordinary valor. His exploits in battle became the stuff of legends, inspiring generations to come. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "3006109", + "images": [ + "3006109_00.jpg", + "3006109_01.jpg", + "3006109_02.jpg", + "3006109_03.jpg", + "3006109_04.jpg", + "3006109_05.jpg", + "3006109_06.jpg", + "3006109_07.jpg", + "3006109_08.jpg", + "3006109_09.jpg", + "3006109_10.jpg", + "3006109_11.jpg", + "3006109_12.jpg", + "3006109_13.jpg", + "3006109_14.jpg", + "3006109_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a group of managers in an office looking out a window celebrating new years \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "1005567", + "images": [ + "1005567_00.jpg", + "1005567_01.jpg", + "1005567_02.jpg", + "1005567_03.jpg", + "1005567_04.jpg", + "1005567_05.jpg", + "1005567_06.jpg", + "1005567_07.jpg", + "1005567_08.jpg", + "1005567_09.jpg", + "1005567_10.jpg", + "1005567_11.jpg", + "1005567_12.jpg", + "1005567_13.jpg", + "1005567_14.jpg", + "1005567_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"old man walking down farm lane at night in a storm \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "6004053", + "images": [ + "6004053_00.jpg", + "6004053_01.jpg", + "6004053_02.jpg", + "6004053_03.jpg", + "6004053_04.jpg", + "6004053_05.jpg", + "6004053_06.jpg", + "6004053_07.jpg", + "6004053_08.jpg", + "6004053_09.jpg", + "6004053_10.jpg", + "6004053_11.jpg", + "6004053_12.jpg", + "6004053_13.jpg", + "6004053_14.jpg", + "6004053_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The first picture shows the moment when a person meets a robot. A person expresses surprise and joy at seeing an intelligent robot standing in front of him, motion 2 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0005888", + "images": [ + "0005888_00.jpg", + "0005888_01.jpg", + "0005888_02.jpg", + "0005888_03.jpg", + "0005888_04.jpg", + "0005888_05.jpg", + "0005888_06.jpg", + "0005888_07.jpg", + "0005888_08.jpg", + "0005888_09.jpg", + "0005888_10.jpg", + "0005888_11.jpg", + "0005888_12.jpg", + "0005888_13.jpg", + "0005888_14.jpg", + "0005888_15.jpg", + "0005888_16.jpg", + "0005888_17.jpg", + "0005888_18.jpg", + "0005888_19.jpg", + "0005888_20.jpg", + "0005888_21.jpg", + "0005888_22.jpg", + "0005888_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A TV showing a scene from Baywatch with david hasselhoff and his tv son hoby, ultra hidef, 32k \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "7003809", + "images": [ + "7003809_00.jpg", + "7003809_01.jpg", + "7003809_02.jpg", + "7003809_03.jpg", + "7003809_04.jpg", + "7003809_05.jpg", + "7003809_06.jpg", + "7003809_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Villagers watching in amazement as the crow solves the problem. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "1004633", + "images": [ + "1004633_00.jpg", + "1004633_01.jpg", + "1004633_02.jpg", + "1004633_03.jpg", + "1004633_04.jpg", + "1004633_05.jpg", + "1004633_06.jpg", + "1004633_07.jpg", + "1004633_08.jpg", + "1004633_09.jpg", + "1004633_10.jpg", + "1004633_11.jpg", + "1004633_12.jpg", + "1004633_13.jpg", + "1004633_14.jpg", + "1004633_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"the hand is floating outside of the clouds, in the style of muted tones, surrealism, blurred imagery, dreamy collages, melting, aerial view \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "5001824", + "images": [ + "5001824_00.jpg", + "5001824_01.jpg", + "5001824_02.jpg", + "5001824_03.jpg", + "5001824_04.jpg", + "5001824_05.jpg", + "5001824_06.jpg", + "5001824_07.jpg", + "5001824_08.jpg", + "5001824_09.jpg", + "5001824_10.jpg", + "5001824_11.jpg", + "5001824_12.jpg", + "5001824_13.jpg", + "5001824_14.jpg", + "5001824_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"create a chart of what would you pick game between 2 things, with 15 different questions. in the first column add the thing1 and in the second column add the thing2. in the third column add an imaginary percentage of how many people would choose thing1 and in the fourth column add an imaginary percentage of how many people would choose thing2.Make it in a tiktok video format. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "3004426", + "images": [ + "3004426_00.jpg", + "3004426_01.jpg", + "3004426_02.jpg", + "3004426_03.jpg", + "3004426_04.jpg", + "3004426_05.jpg", + "3004426_06.jpg", + "3004426_07.jpg", + "3004426_08.jpg", + "3004426_09.jpg", + "3004426_10.jpg", + "3004426_11.jpg", + "3004426_12.jpg", + "3004426_13.jpg", + "3004426_14.jpg", + "3004426_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"winwood walls style image with movement \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "2003255", + "images": [ + "2003255_00.jpg", + "2003255_01.jpg", + "2003255_02.jpg", + "2003255_03.jpg", + "2003255_04.jpg", + "2003255_05.jpg", + "2003255_06.jpg", + "2003255_07.jpg", + "2003255_08.jpg", + "2003255_09.jpg", + "2003255_10.jpg", + "2003255_11.jpg", + "2003255_12.jpg", + "2003255_13.jpg", + "2003255_14.jpg", + "2003255_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"danceing cyborg cats in a high tech space disco, realistic, cinematic, 50 mm, DOF, smooth rythmatic motion, looped \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "4004564", + "images": [ + "4004564_00.jpg", + "4004564_01.jpg", + "4004564_02.jpg", + "4004564_03.jpg", + "4004564_04.jpg", + "4004564_05.jpg", + "4004564_06.jpg", + "4004564_07.jpg", + "4004564_08.jpg", + "4004564_09.jpg", + "4004564_10.jpg", + "4004564_11.jpg", + "4004564_12.jpg", + "4004564_13.jpg", + "4004564_14.jpg", + "4004564_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"animation of a treasure chest opening with gold, glitter, diamonds \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "1003649", + "images": [ + "1003649_00.jpg", + "1003649_01.jpg", + "1003649_02.jpg", + "1003649_03.jpg", + "1003649_04.jpg", + "1003649_05.jpg", + "1003649_06.jpg", + "1003649_07.jpg", + "1003649_08.jpg", + "1003649_09.jpg", + "1003649_10.jpg", + "1003649_11.jpg", + "1003649_12.jpg", + "1003649_13.jpg", + "1003649_14.jpg", + "1003649_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Ukrainian military defends itself against Russian occupiers \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2003991", + "images": [ + "2003991_00.jpg", + "2003991_01.jpg", + "2003991_02.jpg", + "2003991_03.jpg", + "2003991_04.jpg", + "2003991_05.jpg", + "2003991_06.jpg", + "2003991_07.jpg", + "2003991_08.jpg", + "2003991_09.jpg", + "2003991_10.jpg", + "2003991_11.jpg", + "2003991_12.jpg", + "2003991_13.jpg", + "2003991_14.jpg", + "2003991_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"in black and white newsreel style, nasa astronauts go bowling on the moon, in the style of black and white 1900s expressionist silent movie \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "0003241", + "images": [ + "0003241_00.jpg", + "0003241_01.jpg", + "0003241_02.jpg", + "0003241_03.jpg", + "0003241_04.jpg", + "0003241_05.jpg", + "0003241_06.jpg", + "0003241_07.jpg", + "0003241_08.jpg", + "0003241_09.jpg", + "0003241_10.jpg", + "0003241_11.jpg", + "0003241_12.jpg", + "0003241_13.jpg", + "0003241_14.jpg", + "0003241_15.jpg", + "0003241_16.jpg", + "0003241_17.jpg", + "0003241_18.jpg", + "0003241_19.jpg", + "0003241_20.jpg", + "0003241_21.jpg", + "0003241_22.jpg", + "0003241_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"her cape is moving, the cow is talking, the cow is moving, the background is moving along with the cow, the mounts are moving away \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "2005169", + "images": [ + "2005169_00.jpg", + "2005169_01.jpg", + "2005169_02.jpg", + "2005169_03.jpg", + "2005169_04.jpg", + "2005169_05.jpg", + "2005169_06.jpg", + "2005169_07.jpg", + "2005169_08.jpg", + "2005169_09.jpg", + "2005169_10.jpg", + "2005169_11.jpg", + "2005169_12.jpg", + "2005169_13.jpg", + "2005169_14.jpg", + "2005169_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The view of mountains and rivers, the endless mountains, the sun is shining high \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "6002299", + "images": [ + "6002299_00.jpg", + "6002299_01.jpg", + "6002299_02.jpg", + "6002299_03.jpg", + "6002299_04.jpg", + "6002299_05.jpg", + "6002299_06.jpg", + "6002299_07.jpg", + "6002299_08.jpg", + "6002299_09.jpg", + "6002299_10.jpg", + "6002299_11.jpg", + "6002299_12.jpg", + "6002299_13.jpg", + "6002299_14.jpg", + "6002299_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"from far aways a dense forest \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003615", + "images": [ + "6003615_00.jpg", + "6003615_01.jpg", + "6003615_02.jpg", + "6003615_03.jpg", + "6003615_04.jpg", + "6003615_05.jpg", + "6003615_06.jpg", + "6003615_07.jpg", + "6003615_08.jpg", + "6003615_09.jpg", + "6003615_10.jpg", + "6003615_11.jpg", + "6003615_12.jpg", + "6003615_13.jpg", + "6003615_14.jpg", + "6003615_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A team of explorers discovering a underwater city \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003009", + "images": [ + "4003009_00.jpg", + "4003009_01.jpg", + "4003009_02.jpg", + "4003009_03.jpg", + "4003009_04.jpg", + "4003009_05.jpg", + "4003009_06.jpg", + "4003009_07.jpg", + "4003009_08.jpg", + "4003009_09.jpg", + "4003009_10.jpg", + "4003009_11.jpg", + "4003009_12.jpg", + "4003009_13.jpg", + "4003009_14.jpg", + "4003009_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Satoshi Kon style. Anime style. Three white light bodies playing in between the clouds of a purple sky. The ground is fluffy clouds. vintage anime style \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "8003316", + "images": [ + "8003316_00.jpg", + "8003316_01.jpg", + "8003316_02.jpg", + "8003316_03.jpg", + "8003316_04.jpg", + "8003316_05.jpg", + "8003316_06.jpg", + "8003316_07.jpg", + "8003316_08.jpg", + "8003316_09.jpg", + "8003316_10.jpg", + "8003316_11.jpg", + "8003316_12.jpg", + "8003316_13.jpg", + "8003316_14.jpg", + "8003316_15.jpg", + "8003316_16.jpg", + "8003316_17.jpg", + "8003316_18.jpg", + "8003316_19.jpg", + "8003316_20.jpg", + "8003316_21.jpg", + "8003316_22.jpg", + "8003316_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"3D animation, fancy breakfast under a cherry blossom tree\",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "2006228", + "images": [ + "2006228_00.jpg", + "2006228_01.jpg", + "2006228_02.jpg", + "2006228_03.jpg", + "2006228_04.jpg", + "2006228_05.jpg", + "2006228_06.jpg", + "2006228_07.jpg", + "2006228_08.jpg", + "2006228_09.jpg", + "2006228_10.jpg", + "2006228_11.jpg", + "2006228_12.jpg", + "2006228_13.jpg", + "2006228_14.jpg", + "2006228_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"headlights shining on the closed area 51 gate at night \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "2006840", + "images": [ + "2006840_00.jpg", + "2006840_01.jpg", + "2006840_02.jpg", + "2006840_03.jpg", + "2006840_04.jpg", + "2006840_05.jpg", + "2006840_06.jpg", + "2006840_07.jpg", + "2006840_08.jpg", + "2006840_09.jpg", + "2006840_10.jpg", + "2006840_11.jpg", + "2006840_12.jpg", + "2006840_13.jpg", + "2006840_14.jpg", + "2006840_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"filling up a plastic cup with milk until it is full and dripping \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "5001438", + "images": [ + "5001438_00.jpg", + "5001438_01.jpg", + "5001438_02.jpg", + "5001438_03.jpg", + "5001438_04.jpg", + "5001438_05.jpg", + "5001438_06.jpg", + "5001438_07.jpg", + "5001438_08.jpg", + "5001438_09.jpg", + "5001438_10.jpg", + "5001438_11.jpg", + "5001438_12.jpg", + "5001438_13.jpg", + "5001438_14.jpg", + "5001438_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Indian king servent who was inside the jail say something. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "1003660", + "images": [ + "1003660_00.jpg", + "1003660_01.jpg", + "1003660_02.jpg", + "1003660_03.jpg", + "1003660_04.jpg", + "1003660_05.jpg", + "1003660_06.jpg", + "1003660_07.jpg", + "1003660_08.jpg", + "1003660_09.jpg", + "1003660_10.jpg", + "1003660_11.jpg", + "1003660_12.jpg", + "1003660_13.jpg", + "1003660_14.jpg", + "1003660_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"pretty woman running in the park high quality \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0005641", + "images": [ + "0005641_00.jpg", + "0005641_01.jpg", + "0005641_02.jpg", + "0005641_03.jpg", + "0005641_04.jpg", + "0005641_05.jpg", + "0005641_06.jpg", + "0005641_07.jpg", + "0005641_08.jpg", + "0005641_09.jpg", + "0005641_10.jpg", + "0005641_11.jpg", + "0005641_12.jpg", + "0005641_13.jpg", + "0005641_14.jpg", + "0005641_15.jpg", + "0005641_16.jpg", + "0005641_17.jpg", + "0005641_18.jpg", + "0005641_19.jpg", + "0005641_20.jpg", + "0005641_21.jpg", + "0005641_22.jpg", + "0005641_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Five, Guzheng music, melodious A woman sits in front of a guzheng, gently plucking the strings with her fingers. Melodious harp, like spring ding dong, like the wind through the bamboo forest. At this moment, the picture presents a landscape painting, as if flowing with the music of the piano.8K,16:9 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "0005184", + "images": [ + "0005184_00.jpg", + "0005184_01.jpg", + "0005184_02.jpg", + "0005184_03.jpg", + "0005184_04.jpg", + "0005184_05.jpg", + "0005184_06.jpg", + "0005184_07.jpg", + "0005184_08.jpg", + "0005184_09.jpg", + "0005184_10.jpg", + "0005184_11.jpg", + "0005184_12.jpg", + "0005184_13.jpg", + "0005184_14.jpg", + "0005184_15.jpg", + "0005184_16.jpg", + "0005184_17.jpg", + "0005184_18.jpg", + "0005184_19.jpg", + "0005184_20.jpg", + "0005184_21.jpg", + "0005184_22.jpg", + "0005184_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Doctor who controlling the tardis with red lipstick pointing at it \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "0003746", + "images": [ + "0003746_00.jpg", + "0003746_01.jpg", + "0003746_02.jpg", + "0003746_03.jpg", + "0003746_04.jpg", + "0003746_05.jpg", + "0003746_06.jpg", + "0003746_07.jpg", + "0003746_08.jpg", + "0003746_09.jpg", + "0003746_10.jpg", + "0003746_11.jpg", + "0003746_12.jpg", + "0003746_13.jpg", + "0003746_14.jpg", + "0003746_15.jpg", + "0003746_16.jpg", + "0003746_17.jpg", + "0003746_18.jpg", + "0003746_19.jpg", + "0003746_20.jpg", + "0003746_21.jpg", + "0003746_22.jpg", + "0003746_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"lazer beam coming from the sky burning cities ar 16:9 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "2006343", + "images": [ + "2006343_00.jpg", + "2006343_01.jpg", + "2006343_02.jpg", + "2006343_03.jpg", + "2006343_04.jpg", + "2006343_05.jpg", + "2006343_06.jpg", + "2006343_07.jpg", + "2006343_08.jpg", + "2006343_09.jpg", + "2006343_10.jpg", + "2006343_11.jpg", + "2006343_12.jpg", + "2006343_13.jpg", + "2006343_14.jpg", + "2006343_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"television styled glitter and stars flying against purple background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "1004653", + "images": [ + "1004653_00.jpg", + "1004653_01.jpg", + "1004653_02.jpg", + "1004653_03.jpg", + "1004653_04.jpg", + "1004653_05.jpg", + "1004653_06.jpg", + "1004653_07.jpg", + "1004653_08.jpg", + "1004653_09.jpg", + "1004653_10.jpg", + "1004653_11.jpg", + "1004653_12.jpg", + "1004653_13.jpg", + "1004653_14.jpg", + "1004653_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"ultra beautiful hd forest with a lot of animals walking \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "8003747", + "images": [ + "8003747_00.jpg", + "8003747_01.jpg", + "8003747_02.jpg", + "8003747_03.jpg", + "8003747_04.jpg", + "8003747_05.jpg", + "8003747_06.jpg", + "8003747_07.jpg", + "8003747_08.jpg", + "8003747_09.jpg", + "8003747_10.jpg", + "8003747_11.jpg", + "8003747_12.jpg", + "8003747_13.jpg", + "8003747_14.jpg", + "8003747_15.jpg", + "8003747_16.jpg", + "8003747_17.jpg", + "8003747_18.jpg", + "8003747_19.jpg", + "8003747_20.jpg", + "8003747_21.jpg", + "8003747_22.jpg", + "8003747_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Giannis Antetokounmpo playing PUBG at a computer \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "7002299", + "images": [ + "7002299_00.jpg", + "7002299_01.jpg", + "7002299_02.jpg", + "7002299_03.jpg", + "7002299_04.jpg", + "7002299_05.jpg", + "7002299_06.jpg", + "7002299_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Albert Einstein is playing as a Santa Claus delivering gifts to pikas \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "2003379", + "images": [ + "2003379_00.jpg", + "2003379_01.jpg", + "2003379_02.jpg", + "2003379_03.jpg", + "2003379_04.jpg", + "2003379_05.jpg", + "2003379_06.jpg", + "2003379_07.jpg", + "2003379_08.jpg", + "2003379_09.jpg", + "2003379_10.jpg", + "2003379_11.jpg", + "2003379_12.jpg", + "2003379_13.jpg", + "2003379_14.jpg", + "2003379_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"An image of the other animals in the forest watching in amazement as Crow saves the fish. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001647", + "images": [ + "8001647_00.jpg", + "8001647_01.jpg", + "8001647_02.jpg", + "8001647_03.jpg", + "8001647_04.jpg", + "8001647_05.jpg", + "8001647_06.jpg", + "8001647_07.jpg", + "8001647_08.jpg", + "8001647_09.jpg", + "8001647_10.jpg", + "8001647_11.jpg", + "8001647_12.jpg", + "8001647_13.jpg", + "8001647_14.jpg", + "8001647_15.jpg", + "8001647_16.jpg", + "8001647_17.jpg", + "8001647_18.jpg", + "8001647_19.jpg", + "8001647_20.jpg", + "8001647_21.jpg", + "8001647_22.jpg", + "8001647_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"generate a white girl wishing a birthday \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "4002276", + "images": [ + "4002276_00.jpg", + "4002276_01.jpg", + "4002276_02.jpg", + "4002276_03.jpg", + "4002276_04.jpg", + "4002276_05.jpg", + "4002276_06.jpg", + "4002276_07.jpg", + "4002276_08.jpg", + "4002276_09.jpg", + "4002276_10.jpg", + "4002276_11.jpg", + "4002276_12.jpg", + "4002276_13.jpg", + "4002276_14.jpg", + "4002276_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"old Italian movie. Man in the hat driving cabriolet in Italian village. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "5001308", + "images": [ + "5001308_00.jpg", + "5001308_01.jpg", + "5001308_02.jpg", + "5001308_03.jpg", + "5001308_04.jpg", + "5001308_05.jpg", + "5001308_06.jpg", + "5001308_07.jpg", + "5001308_08.jpg", + "5001308_09.jpg", + "5001308_10.jpg", + "5001308_11.jpg", + "5001308_12.jpg", + "5001308_13.jpg", + "5001308_14.jpg", + "5001308_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A little fairy talking to kids \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "3004368", + "images": [ + "3004368_00.jpg", + "3004368_01.jpg", + "3004368_02.jpg", + "3004368_03.jpg", + "3004368_04.jpg", + "3004368_05.jpg", + "3004368_06.jpg", + "3004368_07.jpg", + "3004368_08.jpg", + "3004368_09.jpg", + "3004368_10.jpg", + "3004368_11.jpg", + "3004368_12.jpg", + "3004368_13.jpg", + "3004368_14.jpg", + "3004368_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Illustrate a bustling market scene, with fresh produce displayed on stalls, attracting villagers eager to purchase. cartoon style \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "3003338", + "images": [ + "3003338_00.jpg", + "3003338_01.jpg", + "3003338_02.jpg", + "3003338_03.jpg", + "3003338_04.jpg", + "3003338_05.jpg", + "3003338_06.jpg", + "3003338_07.jpg", + "3003338_08.jpg", + "3003338_09.jpg", + "3003338_10.jpg", + "3003338_11.jpg", + "3003338_12.jpg", + "3003338_13.jpg", + "3003338_14.jpg", + "3003338_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"exhausted man 20s looking up at the rain against blurred neon background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "2003138", + "images": [ + "2003138_00.jpg", + "2003138_01.jpg", + "2003138_02.jpg", + "2003138_03.jpg", + "2003138_04.jpg", + "2003138_05.jpg", + "2003138_06.jpg", + "2003138_07.jpg", + "2003138_08.jpg", + "2003138_09.jpg", + "2003138_10.jpg", + "2003138_11.jpg", + "2003138_12.jpg", + "2003138_13.jpg", + "2003138_14.jpg", + "2003138_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The DOTA2 hero Alchemist is holding a shining Radiance in his hand. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "1004909", + "images": [ + "1004909_00.jpg", + "1004909_01.jpg", + "1004909_02.jpg", + "1004909_03.jpg", + "1004909_04.jpg", + "1004909_05.jpg", + "1004909_06.jpg", + "1004909_07.jpg", + "1004909_08.jpg", + "1004909_09.jpg", + "1004909_10.jpg", + "1004909_11.jpg", + "1004909_12.jpg", + "1004909_13.jpg", + "1004909_14.jpg", + "1004909_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a wolf jumping onto an outstretched arm and transforming into a shield \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "2005543", + "images": [ + "2005543_00.jpg", + "2005543_01.jpg", + "2005543_02.jpg", + "2005543_03.jpg", + "2005543_04.jpg", + "2005543_05.jpg", + "2005543_06.jpg", + "2005543_07.jpg", + "2005543_08.jpg", + "2005543_09.jpg", + "2005543_10.jpg", + "2005543_11.jpg", + "2005543_12.jpg", + "2005543_13.jpg", + "2005543_14.jpg", + "2005543_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Behind it, however, was the suffering of many thousands of farmers who lost their livelihoods and became destitute. It was a serious matter in the end, \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003249", + "images": [ + "4003249_00.jpg", + "4003249_01.jpg", + "4003249_02.jpg", + "4003249_03.jpg", + "4003249_04.jpg", + "4003249_05.jpg", + "4003249_06.jpg", + "4003249_07.jpg", + "4003249_08.jpg", + "4003249_09.jpg", + "4003249_10.jpg", + "4003249_11.jpg", + "4003249_12.jpg", + "4003249_13.jpg", + "4003249_14.jpg", + "4003249_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a person exploring breathtaking virtual landscapes as an explorer avatar \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "8003735", + "images": [ + "8003735_00.jpg", + "8003735_01.jpg", + "8003735_02.jpg", + "8003735_03.jpg", + "8003735_04.jpg", + "8003735_05.jpg", + "8003735_06.jpg", + "8003735_07.jpg", + "8003735_08.jpg", + "8003735_09.jpg", + "8003735_10.jpg", + "8003735_11.jpg", + "8003735_12.jpg", + "8003735_13.jpg", + "8003735_14.jpg", + "8003735_15.jpg", + "8003735_16.jpg", + "8003735_17.jpg", + "8003735_18.jpg", + "8003735_19.jpg", + "8003735_20.jpg", + "8003735_21.jpg", + "8003735_22.jpg", + "8003735_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The video ending with Lily and her friends waving goodbye to the audience. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "8003934", + "images": [ + "8003934_00.jpg", + "8003934_01.jpg", + "8003934_02.jpg", + "8003934_03.jpg", + "8003934_04.jpg", + "8003934_05.jpg", + "8003934_06.jpg", + "8003934_07.jpg", + "8003934_08.jpg", + "8003934_09.jpg", + "8003934_10.jpg", + "8003934_11.jpg", + "8003934_12.jpg", + "8003934_13.jpg", + "8003934_14.jpg", + "8003934_15.jpg", + "8003934_16.jpg", + "8003934_17.jpg", + "8003934_18.jpg", + "8003934_19.jpg", + "8003934_20.jpg", + "8003934_21.jpg", + "8003934_22.jpg", + "8003934_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Imagine her fingers, nimble and determined, dancing across the keyboard like a choreographed performance, the rhythmic tapping echoing in the hushed library. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "1005293", + "images": [ + "1005293_00.jpg", + "1005293_01.jpg", + "1005293_02.jpg", + "1005293_03.jpg", + "1005293_04.jpg", + "1005293_05.jpg", + "1005293_06.jpg", + "1005293_07.jpg", + "1005293_08.jpg", + "1005293_09.jpg", + "1005293_10.jpg", + "1005293_11.jpg", + "1005293_12.jpg", + "1005293_13.jpg", + "1005293_14.jpg", + "1005293_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A Chinese girl, 18 years old, standing on the street, expression of surprise, frontal view, night, frontal view, close up, horror scene, eerie, dim, 8K, graphic novel sketch, graphic novel style, 2D effect, flat color, flat color, gloomy painting style, Horror Thriller,mystery novel \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "3005546", + "images": [ + "3005546_00.jpg", + "3005546_01.jpg", + "3005546_02.jpg", + "3005546_03.jpg", + "3005546_04.jpg", + "3005546_05.jpg", + "3005546_06.jpg", + "3005546_07.jpg", + "3005546_08.jpg", + "3005546_09.jpg", + "3005546_10.jpg", + "3005546_11.jpg", + "3005546_12.jpg", + "3005546_13.jpg", + "3005546_14.jpg", + "3005546_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"maybe with a little knapsack for conveying. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6004601", + "images": [ + "6004601_00.jpg", + "6004601_01.jpg", + "6004601_02.jpg", + "6004601_03.jpg", + "6004601_04.jpg", + "6004601_05.jpg", + "6004601_06.jpg", + "6004601_07.jpg", + "6004601_08.jpg", + "6004601_09.jpg", + "6004601_10.jpg", + "6004601_11.jpg", + "6004601_12.jpg", + "6004601_13.jpg", + "6004601_14.jpg", + "6004601_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A child sits in a car and the car drives away \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0003285", + "images": [ + "0003285_00.jpg", + "0003285_01.jpg", + "0003285_02.jpg", + "0003285_03.jpg", + "0003285_04.jpg", + "0003285_05.jpg", + "0003285_06.jpg", + "0003285_07.jpg", + "0003285_08.jpg", + "0003285_09.jpg", + "0003285_10.jpg", + "0003285_11.jpg", + "0003285_12.jpg", + "0003285_13.jpg", + "0003285_14.jpg", + "0003285_15.jpg", + "0003285_16.jpg", + "0003285_17.jpg", + "0003285_18.jpg", + "0003285_19.jpg", + "0003285_20.jpg", + "0003285_21.jpg", + "0003285_22.jpg", + "0003285_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"shapes of human walking out of a large cave filled with fire \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0004983", + "images": [ + "0004983_00.jpg", + "0004983_01.jpg", + "0004983_02.jpg", + "0004983_03.jpg", + "0004983_04.jpg", + "0004983_05.jpg", + "0004983_06.jpg", + "0004983_07.jpg", + "0004983_08.jpg", + "0004983_09.jpg", + "0004983_10.jpg", + "0004983_11.jpg", + "0004983_12.jpg", + "0004983_13.jpg", + "0004983_14.jpg", + "0004983_15.jpg", + "0004983_16.jpg", + "0004983_17.jpg", + "0004983_18.jpg", + "0004983_19.jpg", + "0004983_20.jpg", + "0004983_21.jpg", + "0004983_22.jpg", + "0004983_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"male rapper in Ethiopia, sitting front of his home, rapping, his girl next to him, 4k, detailed \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "1006820", + "images": [ + "1006820_00.jpg", + "1006820_01.jpg", + "1006820_02.jpg", + "1006820_03.jpg", + "1006820_04.jpg", + "1006820_05.jpg", + "1006820_06.jpg", + "1006820_07.jpg", + "1006820_08.jpg", + "1006820_09.jpg", + "1006820_10.jpg", + "1006820_11.jpg", + "1006820_12.jpg", + "1006820_13.jpg", + "1006820_14.jpg", + "1006820_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"futuristic city in the netherlands containing solar parks, wind turbines in a green surrounding \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0006144", + "images": [ + "0006144_00.jpg", + "0006144_01.jpg", + "0006144_02.jpg", + "0006144_03.jpg", + "0006144_04.jpg", + "0006144_05.jpg", + "0006144_06.jpg", + "0006144_07.jpg", + "0006144_08.jpg", + "0006144_09.jpg", + "0006144_10.jpg", + "0006144_11.jpg", + "0006144_12.jpg", + "0006144_13.jpg", + "0006144_14.jpg", + "0006144_15.jpg", + "0006144_16.jpg", + "0006144_17.jpg", + "0006144_18.jpg", + "0006144_19.jpg", + "0006144_20.jpg", + "0006144_21.jpg", + "0006144_22.jpg", + "0006144_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"An aerial drone shot of a lonely cabin on a snow capped mountain. Cinematic, ultra realistic, shallow focus, 35mm \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "0003566", + "images": [ + "0003566_00.jpg", + "0003566_01.jpg", + "0003566_02.jpg", + "0003566_03.jpg", + "0003566_04.jpg", + "0003566_05.jpg", + "0003566_06.jpg", + "0003566_07.jpg", + "0003566_08.jpg", + "0003566_09.jpg", + "0003566_10.jpg", + "0003566_11.jpg", + "0003566_12.jpg", + "0003566_13.jpg", + "0003566_14.jpg", + "0003566_15.jpg", + "0003566_16.jpg", + "0003566_17.jpg", + "0003566_18.jpg", + "0003566_19.jpg", + "0003566_20.jpg", + "0003566_21.jpg", + "0003566_22.jpg", + "0003566_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a steam train coming towards the screen Alfred Stieglitz grainy black and white 1900s 4k \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "6002018", + "images": [ + "6002018_00.jpg", + "6002018_01.jpg", + "6002018_02.jpg", + "6002018_03.jpg", + "6002018_04.jpg", + "6002018_05.jpg", + "6002018_06.jpg", + "6002018_07.jpg", + "6002018_08.jpg", + "6002018_09.jpg", + "6002018_10.jpg", + "6002018_11.jpg", + "6002018_12.jpg", + "6002018_13.jpg", + "6002018_14.jpg", + "6002018_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"chalk drawing of a scorpion coming to life \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "0006570", + "images": [ + "0006570_00.jpg", + "0006570_01.jpg", + "0006570_02.jpg", + "0006570_03.jpg", + "0006570_04.jpg", + "0006570_05.jpg", + "0006570_06.jpg", + "0006570_07.jpg", + "0006570_08.jpg", + "0006570_09.jpg", + "0006570_10.jpg", + "0006570_11.jpg", + "0006570_12.jpg", + "0006570_13.jpg", + "0006570_14.jpg", + "0006570_15.jpg", + "0006570_16.jpg", + "0006570_17.jpg", + "0006570_18.jpg", + "0006570_19.jpg", + "0006570_20.jpg", + "0006570_21.jpg", + "0006570_22.jpg", + "0006570_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"logo with letters: covers shrift: comic sans \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "6002366", + "images": [ + "6002366_00.jpg", + "6002366_01.jpg", + "6002366_02.jpg", + "6002366_03.jpg", + "6002366_04.jpg", + "6002366_05.jpg", + "6002366_06.jpg", + "6002366_07.jpg", + "6002366_08.jpg", + "6002366_09.jpg", + "6002366_10.jpg", + "6002366_11.jpg", + "6002366_12.jpg", + "6002366_13.jpg", + "6002366_14.jpg", + "6002366_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"flying colourful parrot landing on a perch \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "2004957", + "images": [ + "2004957_00.jpg", + "2004957_01.jpg", + "2004957_02.jpg", + "2004957_03.jpg", + "2004957_04.jpg", + "2004957_05.jpg", + "2004957_06.jpg", + "2004957_07.jpg", + "2004957_08.jpg", + "2004957_09.jpg", + "2004957_10.jpg", + "2004957_11.jpg", + "2004957_12.jpg", + "2004957_13.jpg", + "2004957_14.jpg", + "2004957_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A dog in an astronaut suit getting on a rocket \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "0005171", + "images": [ + "0005171_00.jpg", + "0005171_01.jpg", + "0005171_02.jpg", + "0005171_03.jpg", + "0005171_04.jpg", + "0005171_05.jpg", + "0005171_06.jpg", + "0005171_07.jpg", + "0005171_08.jpg", + "0005171_09.jpg", + "0005171_10.jpg", + "0005171_11.jpg", + "0005171_12.jpg", + "0005171_13.jpg", + "0005171_14.jpg", + "0005171_15.jpg", + "0005171_16.jpg", + "0005171_17.jpg", + "0005171_18.jpg", + "0005171_19.jpg", + "0005171_20.jpg", + "0005171_21.jpg", + "0005171_22.jpg", + "0005171_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a girl who makes a video call \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "1004434", + "images": [ + "1004434_00.jpg", + "1004434_01.jpg", + "1004434_02.jpg", + "1004434_03.jpg", + "1004434_04.jpg", + "1004434_05.jpg", + "1004434_06.jpg", + "1004434_07.jpg", + "1004434_08.jpg", + "1004434_09.jpg", + "1004434_10.jpg", + "1004434_11.jpg", + "1004434_12.jpg", + "1004434_13.jpg", + "1004434_14.jpg", + "1004434_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A robot walking on the moon, filled with stars and sparkling meteors \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "4004766", + "images": [ + "4004766_00.jpg", + "4004766_01.jpg", + "4004766_02.jpg", + "4004766_03.jpg", + "4004766_04.jpg", + "4004766_05.jpg", + "4004766_06.jpg", + "4004766_07.jpg", + "4004766_08.jpg", + "4004766_09.jpg", + "4004766_10.jpg", + "4004766_11.jpg", + "4004766_12.jpg", + "4004766_13.jpg", + "4004766_14.jpg", + "4004766_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cyber books flying in zero gravity, yellow and black color scheme \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "2006901", + "images": [ + "2006901_00.jpg", + "2006901_01.jpg", + "2006901_02.jpg", + "2006901_03.jpg", + "2006901_04.jpg", + "2006901_05.jpg", + "2006901_06.jpg", + "2006901_07.jpg", + "2006901_08.jpg", + "2006901_09.jpg", + "2006901_10.jpg", + "2006901_11.jpg", + "2006901_12.jpg", + "2006901_13.jpg", + "2006901_14.jpg", + "2006901_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"in school, The Insurance trainer explains the main points of car insurance reform \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "3004929", + "images": [ + "3004929_00.jpg", + "3004929_01.jpg", + "3004929_02.jpg", + "3004929_03.jpg", + "3004929_04.jpg", + "3004929_05.jpg", + "3004929_06.jpg", + "3004929_07.jpg", + "3004929_08.jpg", + "3004929_09.jpg", + "3004929_10.jpg", + "3004929_11.jpg", + "3004929_12.jpg", + "3004929_13.jpg", + "3004929_14.jpg", + "3004929_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"the man flying slowly and too clouds \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004059", + "images": [ + "0004059_00.jpg", + "0004059_01.jpg", + "0004059_02.jpg", + "0004059_03.jpg", + "0004059_04.jpg", + "0004059_05.jpg", + "0004059_06.jpg", + "0004059_07.jpg", + "0004059_08.jpg", + "0004059_09.jpg", + "0004059_10.jpg", + "0004059_11.jpg", + "0004059_12.jpg", + "0004059_13.jpg", + "0004059_14.jpg", + "0004059_15.jpg", + "0004059_16.jpg", + "0004059_17.jpg", + "0004059_18.jpg", + "0004059_19.jpg", + "0004059_20.jpg", + "0004059_21.jpg", + "0004059_22.jpg", + "0004059_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"create a video of two boys agruing with each other in a Paris street \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2004169", + "images": [ + "2004169_00.jpg", + "2004169_01.jpg", + "2004169_02.jpg", + "2004169_03.jpg", + "2004169_04.jpg", + "2004169_05.jpg", + "2004169_06.jpg", + "2004169_07.jpg", + "2004169_08.jpg", + "2004169_09.jpg", + "2004169_10.jpg", + "2004169_11.jpg", + "2004169_12.jpg", + "2004169_13.jpg", + "2004169_14.jpg", + "2004169_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"court of law with everything technological and holograms, hyperrealistic \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "3006470", + "images": [ + "3006470_00.jpg", + "3006470_01.jpg", + "3006470_02.jpg", + "3006470_03.jpg", + "3006470_04.jpg", + "3006470_05.jpg", + "3006470_06.jpg", + "3006470_07.jpg", + "3006470_08.jpg", + "3006470_09.jpg", + "3006470_10.jpg", + "3006470_11.jpg", + "3006470_12.jpg", + "3006470_13.jpg", + "3006470_14.jpg", + "3006470_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"One day, while hiking across the icy terrain, they stumbled upon a pack of wolves like none they had ever seen. These were no ordinary wolves. They were larger, more aggressive, and had strange, eerie markings on their fur. These mutant wolves had adapted to the harsh Siberian environment, becoming formidable predators. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7003976", + "images": [ + "7003976_00.jpg", + "7003976_01.jpg", + "7003976_02.jpg", + "7003976_03.jpg", + "7003976_04.jpg", + "7003976_05.jpg", + "7003976_06.jpg", + "7003976_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Out of the temple came the seven angels with the seven plagues. They were dressed in clean, shining linen and wore golden sashes around their chests \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "7003498", + "images": [ + "7003498_00.jpg", + "7003498_01.jpg", + "7003498_02.jpg", + "7003498_03.jpg", + "7003498_04.jpg", + "7003498_05.jpg", + "7003498_06.jpg", + "7003498_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"bentley continental gt driving on the way towards his mansion on the hills \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "7004879", + "images": [ + "7004879_00.jpg", + "7004879_01.jpg", + "7004879_02.jpg", + "7004879_03.jpg", + "7004879_04.jpg", + "7004879_05.jpg", + "7004879_06.jpg", + "7004879_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"User and AI process huge data streams with interface \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "1003493", + "images": [ + "1003493_00.jpg", + "1003493_01.jpg", + "1003493_02.jpg", + "1003493_03.jpg", + "1003493_04.jpg", + "1003493_05.jpg", + "1003493_06.jpg", + "1003493_07.jpg", + "1003493_08.jpg", + "1003493_09.jpg", + "1003493_10.jpg", + "1003493_11.jpg", + "1003493_12.jpg", + "1003493_13.jpg", + "1003493_14.jpg", + "1003493_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"\",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "1005945", + "images": [ + "1005945_00.jpg", + "1005945_01.jpg", + "1005945_02.jpg", + "1005945_03.jpg", + "1005945_04.jpg", + "1005945_05.jpg", + "1005945_06.jpg", + "1005945_07.jpg", + "1005945_08.jpg", + "1005945_09.jpg", + "1005945_10.jpg", + "1005945_11.jpg", + "1005945_12.jpg", + "1005945_13.jpg", + "1005945_14.jpg", + "1005945_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"car emblem shines at golden hour, circle gimbal shot \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "2003440", + "images": [ + "2003440_00.jpg", + "2003440_01.jpg", + "2003440_02.jpg", + "2003440_03.jpg", + "2003440_04.jpg", + "2003440_05.jpg", + "2003440_06.jpg", + "2003440_07.jpg", + "2003440_08.jpg", + "2003440_09.jpg", + "2003440_10.jpg", + "2003440_11.jpg", + "2003440_12.jpg", + "2003440_13.jpg", + "2003440_14.jpg", + "2003440_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"in a magic world, a man find a magic book and find some clues \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "6004540", + "images": [ + "6004540_00.jpg", + "6004540_01.jpg", + "6004540_02.jpg", + "6004540_03.jpg", + "6004540_04.jpg", + "6004540_05.jpg", + "6004540_06.jpg", + "6004540_07.jpg", + "6004540_08.jpg", + "6004540_09.jpg", + "6004540_10.jpg", + "6004540_11.jpg", + "6004540_12.jpg", + "6004540_13.jpg", + "6004540_14.jpg", + "6004540_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"high quality, highly detailed, 8K Ultra HD, detailed illustration of the interior of a perfect forest with mysterious constellations and beautifully designed cosmic spirals, fabric art, art station, splash arts, luminism, 3D rendering, octane rendering, Isometric, by yukisakura, incredible full color, \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003240", + "images": [ + "4003240_00.jpg", + "4003240_01.jpg", + "4003240_02.jpg", + "4003240_03.jpg", + "4003240_04.jpg", + "4003240_05.jpg", + "4003240_06.jpg", + "4003240_07.jpg", + "4003240_08.jpg", + "4003240_09.jpg", + "4003240_10.jpg", + "4003240_11.jpg", + "4003240_12.jpg", + "4003240_13.jpg", + "4003240_14.jpg", + "4003240_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A female ghost dressed in red wedding attire sits atop a grave mound, feasting on the offerings, while the worshipper is scared a lot \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "1004776", + "images": [ + "1004776_00.jpg", + "1004776_01.jpg", + "1004776_02.jpg", + "1004776_03.jpg", + "1004776_04.jpg", + "1004776_05.jpg", + "1004776_06.jpg", + "1004776_07.jpg", + "1004776_08.jpg", + "1004776_09.jpg", + "1004776_10.jpg", + "1004776_11.jpg", + "1004776_12.jpg", + "1004776_13.jpg", + "1004776_14.jpg", + "1004776_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"the parents are fighting, the child in the middle is crying \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "8003897", + "images": [ + "8003897_00.jpg", + "8003897_01.jpg", + "8003897_02.jpg", + "8003897_03.jpg", + "8003897_04.jpg", + "8003897_05.jpg", + "8003897_06.jpg", + "8003897_07.jpg", + "8003897_08.jpg", + "8003897_09.jpg", + "8003897_10.jpg", + "8003897_11.jpg", + "8003897_12.jpg", + "8003897_13.jpg", + "8003897_14.jpg", + "8003897_15.jpg", + "8003897_16.jpg", + "8003897_17.jpg", + "8003897_18.jpg", + "8003897_19.jpg", + "8003897_20.jpg", + "8003897_21.jpg", + "8003897_22.jpg", + "8003897_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"nice big church,in the background people are talking and a demon is peeping from inside the church \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002493", + "images": [ + "7002493_00.jpg", + "7002493_01.jpg", + "7002493_02.jpg", + "7002493_03.jpg", + "7002493_04.jpg", + "7002493_05.jpg", + "7002493_06.jpg", + "7002493_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a majestic kangaroo jumping a long distance, realistic look. 9:16 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "6002475", + "images": [ + "6002475_00.jpg", + "6002475_01.jpg", + "6002475_02.jpg", + "6002475_03.jpg", + "6002475_04.jpg", + "6002475_05.jpg", + "6002475_06.jpg", + "6002475_07.jpg", + "6002475_08.jpg", + "6002475_09.jpg", + "6002475_10.jpg", + "6002475_11.jpg", + "6002475_12.jpg", + "6002475_13.jpg", + "6002475_14.jpg", + "6002475_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"characters from league of legends run \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "6003225", + "images": [ + "6003225_00.jpg", + "6003225_01.jpg", + "6003225_02.jpg", + "6003225_03.jpg", + "6003225_04.jpg", + "6003225_05.jpg", + "6003225_06.jpg", + "6003225_07.jpg", + "6003225_08.jpg", + "6003225_09.jpg", + "6003225_10.jpg", + "6003225_11.jpg", + "6003225_12.jpg", + "6003225_13.jpg", + "6003225_14.jpg", + "6003225_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Draw a ghost wandering in the office \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "1003862", + "images": [ + "1003862_00.jpg", + "1003862_01.jpg", + "1003862_02.jpg", + "1003862_03.jpg", + "1003862_04.jpg", + "1003862_05.jpg", + "1003862_06.jpg", + "1003862_07.jpg", + "1003862_08.jpg", + "1003862_09.jpg", + "1003862_10.jpg", + "1003862_11.jpg", + "1003862_12.jpg", + "1003862_13.jpg", + "1003862_14.jpg", + "1003862_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a robot running through a new york like futuristic city, ultra realistic \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3003318", + "images": [ + "3003318_00.jpg", + "3003318_01.jpg", + "3003318_02.jpg", + "3003318_03.jpg", + "3003318_04.jpg", + "3003318_05.jpg", + "3003318_06.jpg", + "3003318_07.jpg", + "3003318_08.jpg", + "3003318_09.jpg", + "3003318_10.jpg", + "3003318_11.jpg", + "3003318_12.jpg", + "3003318_13.jpg", + "3003318_14.jpg", + "3003318_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a panda in a suit speaks to the camera \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "3006378", + "images": [ + "3006378_00.jpg", + "3006378_01.jpg", + "3006378_02.jpg", + "3006378_03.jpg", + "3006378_04.jpg", + "3006378_05.jpg", + "3006378_06.jpg", + "3006378_07.jpg", + "3006378_08.jpg", + "3006378_09.jpg", + "3006378_10.jpg", + "3006378_11.jpg", + "3006378_12.jpg", + "3006378_13.jpg", + "3006378_14.jpg", + "3006378_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Six wizards stand in a defensive formation against an army of werewolves, they are surrounded on all sides and it is snowing heavily. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "5001820", + "images": [ + "5001820_00.jpg", + "5001820_01.jpg", + "5001820_02.jpg", + "5001820_03.jpg", + "5001820_04.jpg", + "5001820_05.jpg", + "5001820_06.jpg", + "5001820_07.jpg", + "5001820_08.jpg", + "5001820_09.jpg", + "5001820_10.jpg", + "5001820_11.jpg", + "5001820_12.jpg", + "5001820_13.jpg", + "5001820_14.jpg", + "5001820_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a cute little fluffy cat walking in a park,3d,anime cartoon \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3004501", + "images": [ + "3004501_00.jpg", + "3004501_01.jpg", + "3004501_02.jpg", + "3004501_03.jpg", + "3004501_04.jpg", + "3004501_05.jpg", + "3004501_06.jpg", + "3004501_07.jpg", + "3004501_08.jpg", + "3004501_09.jpg", + "3004501_10.jpg", + "3004501_11.jpg", + "3004501_12.jpg", + "3004501_13.jpg", + "3004501_14.jpg", + "3004501_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a crow flying and sitting on a pot, ground is green sky is blue and light coming from the east, the crow is also flying from the east \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "3006650", + "images": [ + "3006650_00.jpg", + "3006650_01.jpg", + "3006650_02.jpg", + "3006650_03.jpg", + "3006650_04.jpg", + "3006650_05.jpg", + "3006650_06.jpg", + "3006650_07.jpg", + "3006650_08.jpg", + "3006650_09.jpg", + "3006650_10.jpg", + "3006650_11.jpg", + "3006650_12.jpg", + "3006650_13.jpg", + "3006650_14.jpg", + "3006650_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a small village with green scenery and evening red sky, night background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "6003277", + "images": [ + "6003277_00.jpg", + "6003277_01.jpg", + "6003277_02.jpg", + "6003277_03.jpg", + "6003277_04.jpg", + "6003277_05.jpg", + "6003277_06.jpg", + "6003277_07.jpg", + "6003277_08.jpg", + "6003277_09.jpg", + "6003277_10.jpg", + "6003277_11.jpg", + "6003277_12.jpg", + "6003277_13.jpg", + "6003277_14.jpg", + "6003277_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Beautiful Cinematic illustration of a laboratory exploding \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "0005117", + "images": [ + "0005117_00.jpg", + "0005117_01.jpg", + "0005117_02.jpg", + "0005117_03.jpg", + "0005117_04.jpg", + "0005117_05.jpg", + "0005117_06.jpg", + "0005117_07.jpg", + "0005117_08.jpg", + "0005117_09.jpg", + "0005117_10.jpg", + "0005117_11.jpg", + "0005117_12.jpg", + "0005117_13.jpg", + "0005117_14.jpg", + "0005117_15.jpg", + "0005117_16.jpg", + "0005117_17.jpg", + "0005117_18.jpg", + "0005117_19.jpg", + "0005117_20.jpg", + "0005117_21.jpg", + "0005117_22.jpg", + "0005117_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Clinging onto the rope, Tom Cruise \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "2004504", + "images": [ + "2004504_00.jpg", + "2004504_01.jpg", + "2004504_02.jpg", + "2004504_03.jpg", + "2004504_04.jpg", + "2004504_05.jpg", + "2004504_06.jpg", + "2004504_07.jpg", + "2004504_08.jpg", + "2004504_09.jpg", + "2004504_10.jpg", + "2004504_11.jpg", + "2004504_12.jpg", + "2004504_13.jpg", + "2004504_14.jpg", + "2004504_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"software company DevTeam offering software solutions for companies \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "0003425", + "images": [ + "0003425_00.jpg", + "0003425_01.jpg", + "0003425_02.jpg", + "0003425_03.jpg", + "0003425_04.jpg", + "0003425_05.jpg", + "0003425_06.jpg", + "0003425_07.jpg", + "0003425_08.jpg", + "0003425_09.jpg", + "0003425_10.jpg", + "0003425_11.jpg", + "0003425_12.jpg", + "0003425_13.jpg", + "0003425_14.jpg", + "0003425_15.jpg", + "0003425_16.jpg", + "0003425_17.jpg", + "0003425_18.jpg", + "0003425_19.jpg", + "0003425_20.jpg", + "0003425_21.jpg", + "0003425_22.jpg", + "0003425_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"field of flowers red yellow white \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "4004604", + "images": [ + "4004604_00.jpg", + "4004604_01.jpg", + "4004604_02.jpg", + "4004604_03.jpg", + "4004604_04.jpg", + "4004604_05.jpg", + "4004604_06.jpg", + "4004604_07.jpg", + "4004604_08.jpg", + "4004604_09.jpg", + "4004604_10.jpg", + "4004604_11.jpg", + "4004604_12.jpg", + "4004604_13.jpg", + "4004604_14.jpg", + "4004604_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Thankfully, a Royal Navy rescue party eventually located and rescued them. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "0006655", + "images": [ + "0006655_00.jpg", + "0006655_01.jpg", + "0006655_02.jpg", + "0006655_03.jpg", + "0006655_04.jpg", + "0006655_05.jpg", + "0006655_06.jpg", + "0006655_07.jpg", + "0006655_08.jpg", + "0006655_09.jpg", + "0006655_10.jpg", + "0006655_11.jpg", + "0006655_12.jpg", + "0006655_13.jpg", + "0006655_14.jpg", + "0006655_15.jpg", + "0006655_16.jpg", + "0006655_17.jpg", + "0006655_18.jpg", + "0006655_19.jpg", + "0006655_20.jpg", + "0006655_21.jpg", + "0006655_22.jpg", + "0006655_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"anime style with lightning in the background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002595", + "images": [ + "7002595_00.jpg", + "7002595_01.jpg", + "7002595_02.jpg", + "7002595_03.jpg", + "7002595_04.jpg", + "7002595_05.jpg", + "7002595_06.jpg", + "7002595_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"women shopping at a really fancy shopping center \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004334", + "images": [ + "0004334_00.jpg", + "0004334_01.jpg", + "0004334_02.jpg", + "0004334_03.jpg", + "0004334_04.jpg", + "0004334_05.jpg", + "0004334_06.jpg", + "0004334_07.jpg", + "0004334_08.jpg", + "0004334_09.jpg", + "0004334_10.jpg", + "0004334_11.jpg", + "0004334_12.jpg", + "0004334_13.jpg", + "0004334_14.jpg", + "0004334_15.jpg", + "0004334_16.jpg", + "0004334_17.jpg", + "0004334_18.jpg", + "0004334_19.jpg", + "0004334_20.jpg", + "0004334_21.jpg", + "0004334_22.jpg", + "0004334_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"man in black coat getting covered in explodion and smoke on street with colorful tenement hauses around, photorealistic 8k \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "5001031", + "images": [ + "5001031_00.jpg", + "5001031_01.jpg", + "5001031_02.jpg", + "5001031_03.jpg", + "5001031_04.jpg", + "5001031_05.jpg", + "5001031_06.jpg", + "5001031_07.jpg", + "5001031_08.jpg", + "5001031_09.jpg", + "5001031_10.jpg", + "5001031_11.jpg", + "5001031_12.jpg", + "5001031_13.jpg", + "5001031_14.jpg", + "5001031_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cinematic, dog and a cat getting married \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "2003969", + "images": [ + "2003969_00.jpg", + "2003969_01.jpg", + "2003969_02.jpg", + "2003969_03.jpg", + "2003969_04.jpg", + "2003969_05.jpg", + "2003969_06.jpg", + "2003969_07.jpg", + "2003969_08.jpg", + "2003969_09.jpg", + "2003969_10.jpg", + "2003969_11.jpg", + "2003969_12.jpg", + "2003969_13.jpg", + "2003969_14.jpg", + "2003969_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"an airplane made of red velvet flies against the background of a sunset over the sea towards the mountains made of blue velvet \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003062", + "images": [ + "4003062_00.jpg", + "4003062_01.jpg", + "4003062_02.jpg", + "4003062_03.jpg", + "4003062_04.jpg", + "4003062_05.jpg", + "4003062_06.jpg", + "4003062_07.jpg", + "4003062_08.jpg", + "4003062_09.jpg", + "4003062_10.jpg", + "4003062_11.jpg", + "4003062_12.jpg", + "4003062_13.jpg", + "4003062_14.jpg", + "4003062_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"video of soldiers walking through mud and the city towards the enemy \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "8001463", + "images": [ + "8001463_00.jpg", + "8001463_01.jpg", + "8001463_02.jpg", + "8001463_03.jpg", + "8001463_04.jpg", + "8001463_05.jpg", + "8001463_06.jpg", + "8001463_07.jpg", + "8001463_08.jpg", + "8001463_09.jpg", + "8001463_10.jpg", + "8001463_11.jpg", + "8001463_12.jpg", + "8001463_13.jpg", + "8001463_14.jpg", + "8001463_15.jpg", + "8001463_16.jpg", + "8001463_17.jpg", + "8001463_18.jpg", + "8001463_19.jpg", + "8001463_20.jpg", + "8001463_21.jpg", + "8001463_22.jpg", + "8001463_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A young and cute Japanese woman is walking happily and lightly on the Togetsukyo Bridge in Arashiyama, Kyoto during late autumn \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0006922", + "images": [ + "0006922_00.jpg", + "0006922_01.jpg", + "0006922_02.jpg", + "0006922_03.jpg", + "0006922_04.jpg", + "0006922_05.jpg", + "0006922_06.jpg", + "0006922_07.jpg", + "0006922_08.jpg", + "0006922_09.jpg", + "0006922_10.jpg", + "0006922_11.jpg", + "0006922_12.jpg", + "0006922_13.jpg", + "0006922_14.jpg", + "0006922_15.jpg", + "0006922_16.jpg", + "0006922_17.jpg", + "0006922_18.jpg", + "0006922_19.jpg", + "0006922_20.jpg", + "0006922_21.jpg", + "0006922_22.jpg", + "0006922_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Joe Biden driving a toyota prius and wear seatbelt \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "7003462", + "images": [ + "7003462_00.jpg", + "7003462_01.jpg", + "7003462_02.jpg", + "7003462_03.jpg", + "7003462_04.jpg", + "7003462_05.jpg", + "7003462_06.jpg", + "7003462_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a plundered temple during 4th century \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "7003684", + "images": [ + "7003684_00.jpg", + "7003684_01.jpg", + "7003684_02.jpg", + "7003684_03.jpg", + "7003684_04.jpg", + "7003684_05.jpg", + "7003684_06.jpg", + "7003684_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a capital L in red on a white potrait back ground being ripped in half \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "2006232", + "images": [ + "2006232_00.jpg", + "2006232_01.jpg", + "2006232_02.jpg", + "2006232_03.jpg", + "2006232_04.jpg", + "2006232_05.jpg", + "2006232_06.jpg", + "2006232_07.jpg", + "2006232_08.jpg", + "2006232_09.jpg", + "2006232_10.jpg", + "2006232_11.jpg", + "2006232_12.jpg", + "2006232_13.jpg", + "2006232_14.jpg", + "2006232_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Profesional photograph of a village by the sea like Isla Cristina, sunset, magic light, 8k, fishing boats, camera zoom out Message: USISA (Font: MODERN) \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "1005646", + "images": [ + "1005646_00.jpg", + "1005646_01.jpg", + "1005646_02.jpg", + "1005646_03.jpg", + "1005646_04.jpg", + "1005646_05.jpg", + "1005646_06.jpg", + "1005646_07.jpg", + "1005646_08.jpg", + "1005646_09.jpg", + "1005646_10.jpg", + "1005646_11.jpg", + "1005646_12.jpg", + "1005646_13.jpg", + "1005646_14.jpg", + "1005646_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a stone floor in a medieval castle hall with a snow storm at night \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "1006001", + "images": [ + "1006001_00.jpg", + "1006001_01.jpg", + "1006001_02.jpg", + "1006001_03.jpg", + "1006001_04.jpg", + "1006001_05.jpg", + "1006001_06.jpg", + "1006001_07.jpg", + "1006001_08.jpg", + "1006001_09.jpg", + "1006001_10.jpg", + "1006001_11.jpg", + "1006001_12.jpg", + "1006001_13.jpg", + "1006001_14.jpg", + "1006001_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"two snakes moves smoothly on the tree of life in Amazon jungle, wide angle, mysterious and magical lightening \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "2003410", + "images": [ + "2003410_00.jpg", + "2003410_01.jpg", + "2003410_02.jpg", + "2003410_03.jpg", + "2003410_04.jpg", + "2003410_05.jpg", + "2003410_06.jpg", + "2003410_07.jpg", + "2003410_08.jpg", + "2003410_09.jpg", + "2003410_10.jpg", + "2003410_11.jpg", + "2003410_12.jpg", + "2003410_13.jpg", + "2003410_14.jpg", + "2003410_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"bright colored yarn being cast on to a massive knitting needle in a suggestive way \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "1004848", + "images": [ + "1004848_00.jpg", + "1004848_01.jpg", + "1004848_02.jpg", + "1004848_03.jpg", + "1004848_04.jpg", + "1004848_05.jpg", + "1004848_06.jpg", + "1004848_07.jpg", + "1004848_08.jpg", + "1004848_09.jpg", + "1004848_10.jpg", + "1004848_11.jpg", + "1004848_12.jpg", + "1004848_13.jpg", + "1004848_14.jpg", + "1004848_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"beautiful anime woman sings for life walking gracefuly slowly along a beach \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0005799", + "images": [ + "0005799_00.jpg", + "0005799_01.jpg", + "0005799_02.jpg", + "0005799_03.jpg", + "0005799_04.jpg", + "0005799_05.jpg", + "0005799_06.jpg", + "0005799_07.jpg", + "0005799_08.jpg", + "0005799_09.jpg", + "0005799_10.jpg", + "0005799_11.jpg", + "0005799_12.jpg", + "0005799_13.jpg", + "0005799_14.jpg", + "0005799_15.jpg", + "0005799_16.jpg", + "0005799_17.jpg", + "0005799_18.jpg", + "0005799_19.jpg", + "0005799_20.jpg", + "0005799_21.jpg", + "0005799_22.jpg", + "0005799_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"gorgeous young woman is posing in white wings, in the style of sculptural paper constructions, mike campau, floral accents, tony cragg, exotic, textured fabrics, angelina wrona \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "5001280", + "images": [ + "5001280_00.jpg", + "5001280_01.jpg", + "5001280_02.jpg", + "5001280_03.jpg", + "5001280_04.jpg", + "5001280_05.jpg", + "5001280_06.jpg", + "5001280_07.jpg", + "5001280_08.jpg", + "5001280_09.jpg", + "5001280_10.jpg", + "5001280_11.jpg", + "5001280_12.jpg", + "5001280_13.jpg", + "5001280_14.jpg", + "5001280_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A flash of lightning illuminates the scene, revealing a path leading through the storm. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "7002235", + "images": [ + "7002235_00.jpg", + "7002235_01.jpg", + "7002235_02.jpg", + "7002235_03.jpg", + "7002235_04.jpg", + "7002235_05.jpg", + "7002235_06.jpg", + "7002235_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"One day, while exploring the rocky shores, she spotted a shimmering tail disappearing beneath the waves) \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7004411", + "images": [ + "7004411_00.jpg", + "7004411_01.jpg", + "7004411_02.jpg", + "7004411_03.jpg", + "7004411_04.jpg", + "7004411_05.jpg", + "7004411_06.jpg", + "7004411_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"mule kneeling at the altar. cinematic, 8k, dark background, natural sunlight. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "3005372", + "images": [ + "3005372_00.jpg", + "3005372_01.jpg", + "3005372_02.jpg", + "3005372_03.jpg", + "3005372_04.jpg", + "3005372_05.jpg", + "3005372_06.jpg", + "3005372_07.jpg", + "3005372_08.jpg", + "3005372_09.jpg", + "3005372_10.jpg", + "3005372_11.jpg", + "3005372_12.jpg", + "3005372_13.jpg", + "3005372_14.jpg", + "3005372_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A woman sleeps in a dark bedroom, sunlight shines through the window. hyper realistic ultra detailed high contrast. Black and white, Film noir style setting, 1940. . 20th centuryfilm nor 1940s \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "8002475", + "images": [ + "8002475_00.jpg", + "8002475_01.jpg", + "8002475_02.jpg", + "8002475_03.jpg", + "8002475_04.jpg", + "8002475_05.jpg", + "8002475_06.jpg", + "8002475_07.jpg", + "8002475_08.jpg", + "8002475_09.jpg", + "8002475_10.jpg", + "8002475_11.jpg", + "8002475_12.jpg", + "8002475_13.jpg", + "8002475_14.jpg", + "8002475_15.jpg", + "8002475_16.jpg", + "8002475_17.jpg", + "8002475_18.jpg", + "8002475_19.jpg", + "8002475_20.jpg", + "8002475_21.jpg", + "8002475_22.jpg", + "8002475_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Porsche is driving through the desert in a storm in the rays of the setting sun and a cave storm is about to overtake him \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "1003751", + "images": [ + "1003751_00.jpg", + "1003751_01.jpg", + "1003751_02.jpg", + "1003751_03.jpg", + "1003751_04.jpg", + "1003751_05.jpg", + "1003751_06.jpg", + "1003751_07.jpg", + "1003751_08.jpg", + "1003751_09.jpg", + "1003751_10.jpg", + "1003751_11.jpg", + "1003751_12.jpg", + "1003751_13.jpg", + "1003751_14.jpg", + "1003751_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"sunlight rays on city on humans \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "2006900", + "images": [ + "2006900_00.jpg", + "2006900_01.jpg", + "2006900_02.jpg", + "2006900_03.jpg", + "2006900_04.jpg", + "2006900_05.jpg", + "2006900_06.jpg", + "2006900_07.jpg", + "2006900_08.jpg", + "2006900_09.jpg", + "2006900_10.jpg", + "2006900_11.jpg", + "2006900_12.jpg", + "2006900_13.jpg", + "2006900_14.jpg", + "2006900_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"give me a reels shorts video like podcasts make a 30sec video \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "2004441", + "images": [ + "2004441_00.jpg", + "2004441_01.jpg", + "2004441_02.jpg", + "2004441_03.jpg", + "2004441_04.jpg", + "2004441_05.jpg", + "2004441_06.jpg", + "2004441_07.jpg", + "2004441_08.jpg", + "2004441_09.jpg", + "2004441_10.jpg", + "2004441_11.jpg", + "2004441_12.jpg", + "2004441_13.jpg", + "2004441_14.jpg", + "2004441_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A Dog, 3D video and 4K quality \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "3006421", + "images": [ + "3006421_00.jpg", + "3006421_01.jpg", + "3006421_02.jpg", + "3006421_03.jpg", + "3006421_04.jpg", + "3006421_05.jpg", + "3006421_06.jpg", + "3006421_07.jpg", + "3006421_08.jpg", + "3006421_09.jpg", + "3006421_10.jpg", + "3006421_11.jpg", + "3006421_12.jpg", + "3006421_13.jpg", + "3006421_14.jpg", + "3006421_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Dark forest with a full moon. Fog rolls in slowly. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "7004871", + "images": [ + "7004871_00.jpg", + "7004871_01.jpg", + "7004871_02.jpg", + "7004871_03.jpg", + "7004871_04.jpg", + "7004871_05.jpg", + "7004871_06.jpg", + "7004871_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a rich guy is shouting for help to save his burning house \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "7002485", + "images": [ + "7002485_00.jpg", + "7002485_01.jpg", + "7002485_02.jpg", + "7002485_03.jpg", + "7002485_04.jpg", + "7002485_05.jpg", + "7002485_06.jpg", + "7002485_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"five nights at freddys location with freddy dancing in front of it \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001688", + "images": [ + "8001688_00.jpg", + "8001688_01.jpg", + "8001688_02.jpg", + "8001688_03.jpg", + "8001688_04.jpg", + "8001688_05.jpg", + "8001688_06.jpg", + "8001688_07.jpg", + "8001688_08.jpg", + "8001688_09.jpg", + "8001688_10.jpg", + "8001688_11.jpg", + "8001688_12.jpg", + "8001688_13.jpg", + "8001688_14.jpg", + "8001688_15.jpg", + "8001688_16.jpg", + "8001688_17.jpg", + "8001688_18.jpg", + "8001688_19.jpg", + "8001688_20.jpg", + "8001688_21.jpg", + "8001688_22.jpg", + "8001688_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cave painting of prehistoric men hunting mammoth animated \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "6002384", + "images": [ + "6002384_00.jpg", + "6002384_01.jpg", + "6002384_02.jpg", + "6002384_03.jpg", + "6002384_04.jpg", + "6002384_05.jpg", + "6002384_06.jpg", + "6002384_07.jpg", + "6002384_08.jpg", + "6002384_09.jpg", + "6002384_10.jpg", + "6002384_11.jpg", + "6002384_12.jpg", + "6002384_13.jpg", + "6002384_14.jpg", + "6002384_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"first person perspective walking inside a closed supermarket with nobody there with dim light \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3003483", + "images": [ + "3003483_00.jpg", + "3003483_01.jpg", + "3003483_02.jpg", + "3003483_03.jpg", + "3003483_04.jpg", + "3003483_05.jpg", + "3003483_06.jpg", + "3003483_07.jpg", + "3003483_08.jpg", + "3003483_09.jpg", + "3003483_10.jpg", + "3003483_11.jpg", + "3003483_12.jpg", + "3003483_13.jpg", + "3003483_14.jpg", + "3003483_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The waterfall rumbled, and the little bear Mumu felt very novel. Suddenly, a group of elves jumped out from behind the tree and among the flowers and leaves. They greeted Mumu the little bear with a smile. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "2004004", + "images": [ + "2004004_00.jpg", + "2004004_01.jpg", + "2004004_02.jpg", + "2004004_03.jpg", + "2004004_04.jpg", + "2004004_05.jpg", + "2004004_06.jpg", + "2004004_07.jpg", + "2004004_08.jpg", + "2004004_09.jpg", + "2004004_10.jpg", + "2004004_11.jpg", + "2004004_12.jpg", + "2004004_13.jpg", + "2004004_14.jpg", + "2004004_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Show a 4k resolution 3D animation of a young talented artist practicing their craft in a small studio. Keywords: Artistry creativity passion dedication. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "2006646", + "images": [ + "2006646_00.jpg", + "2006646_01.jpg", + "2006646_02.jpg", + "2006646_03.jpg", + "2006646_04.jpg", + "2006646_05.jpg", + "2006646_06.jpg", + "2006646_07.jpg", + "2006646_08.jpg", + "2006646_09.jpg", + "2006646_10.jpg", + "2006646_11.jpg", + "2006646_12.jpg", + "2006646_13.jpg", + "2006646_14.jpg", + "2006646_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"zoom in to the centre of the picture \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003625", + "images": [ + "6003625_00.jpg", + "6003625_01.jpg", + "6003625_02.jpg", + "6003625_03.jpg", + "6003625_04.jpg", + "6003625_05.jpg", + "6003625_06.jpg", + "6003625_07.jpg", + "6003625_08.jpg", + "6003625_09.jpg", + "6003625_10.jpg", + "6003625_11.jpg", + "6003625_12.jpg", + "6003625_13.jpg", + "6003625_14.jpg", + "6003625_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A man sits in a dark room near an open window and smokes a cigarette \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6002337", + "images": [ + "6002337_00.jpg", + "6002337_01.jpg", + "6002337_02.jpg", + "6002337_03.jpg", + "6002337_04.jpg", + "6002337_05.jpg", + "6002337_06.jpg", + "6002337_07.jpg", + "6002337_08.jpg", + "6002337_09.jpg", + "6002337_10.jpg", + "6002337_11.jpg", + "6002337_12.jpg", + "6002337_13.jpg", + "6002337_14.jpg", + "6002337_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a top down drone view of a busy london street in 1990 \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7002495", + "images": [ + "7002495_00.jpg", + "7002495_01.jpg", + "7002495_02.jpg", + "7002495_03.jpg", + "7002495_04.jpg", + "7002495_05.jpg", + "7002495_06.jpg", + "7002495_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Generate an hyper realistic depiction of the Battle of Gaugamela, showcasing the intensity and scale of the conflict,full hd \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "0004216", + "images": [ + "0004216_00.jpg", + "0004216_01.jpg", + "0004216_02.jpg", + "0004216_03.jpg", + "0004216_04.jpg", + "0004216_05.jpg", + "0004216_06.jpg", + "0004216_07.jpg", + "0004216_08.jpg", + "0004216_09.jpg", + "0004216_10.jpg", + "0004216_11.jpg", + "0004216_12.jpg", + "0004216_13.jpg", + "0004216_14.jpg", + "0004216_15.jpg", + "0004216_16.jpg", + "0004216_17.jpg", + "0004216_18.jpg", + "0004216_19.jpg", + "0004216_20.jpg", + "0004216_21.jpg", + "0004216_22.jpg", + "0004216_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"half hummingbird, half rabbit, drinking out of a river with the Eiffel tower in the background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "2003670", + "images": [ + "2003670_00.jpg", + "2003670_01.jpg", + "2003670_02.jpg", + "2003670_03.jpg", + "2003670_04.jpg", + "2003670_05.jpg", + "2003670_06.jpg", + "2003670_07.jpg", + "2003670_08.jpg", + "2003670_09.jpg", + "2003670_10.jpg", + "2003670_11.jpg", + "2003670_12.jpg", + "2003670_13.jpg", + "2003670_14.jpg", + "2003670_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"create a beautiful young boy wearing cap on his head nicely eating food nicely sitting on a chair with green screen background \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "3003746", + "images": [ + "3003746_00.jpg", + "3003746_01.jpg", + "3003746_02.jpg", + "3003746_03.jpg", + "3003746_04.jpg", + "3003746_05.jpg", + "3003746_06.jpg", + "3003746_07.jpg", + "3003746_08.jpg", + "3003746_09.jpg", + "3003746_10.jpg", + "3003746_11.jpg", + "3003746_12.jpg", + "3003746_13.jpg", + "3003746_14.jpg", + "3003746_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"girl standing still, penguin move around,wave in sea , \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "8002896", + "images": [ + "8002896_00.jpg", + "8002896_01.jpg", + "8002896_02.jpg", + "8002896_03.jpg", + "8002896_04.jpg", + "8002896_05.jpg", + "8002896_06.jpg", + "8002896_07.jpg", + "8002896_08.jpg", + "8002896_09.jpg", + "8002896_10.jpg", + "8002896_11.jpg", + "8002896_12.jpg", + "8002896_13.jpg", + "8002896_14.jpg", + "8002896_15.jpg", + "8002896_16.jpg", + "8002896_17.jpg", + "8002896_18.jpg", + "8002896_19.jpg", + "8002896_20.jpg", + "8002896_21.jpg", + "8002896_22.jpg", + "8002896_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"boeing 777 300er cruising above atlantic ocean during the sunset \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003909", + "images": [ + "4003909_00.jpg", + "4003909_01.jpg", + "4003909_02.jpg", + "4003909_03.jpg", + "4003909_04.jpg", + "4003909_05.jpg", + "4003909_06.jpg", + "4003909_07.jpg", + "4003909_08.jpg", + "4003909_09.jpg", + "4003909_10.jpg", + "4003909_11.jpg", + "4003909_12.jpg", + "4003909_13.jpg", + "4003909_14.jpg", + "4003909_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The Lich King from the game World of Warcraft without armor \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "6004069", + "images": [ + "6004069_00.jpg", + "6004069_01.jpg", + "6004069_02.jpg", + "6004069_03.jpg", + "6004069_04.jpg", + "6004069_05.jpg", + "6004069_06.jpg", + "6004069_07.jpg", + "6004069_08.jpg", + "6004069_09.jpg", + "6004069_10.jpg", + "6004069_11.jpg", + "6004069_12.jpg", + "6004069_13.jpg", + "6004069_14.jpg", + "6004069_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a man in a white suit driving a customized jeep chasing a bison across the plains \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "4004321", + "images": [ + "4004321_00.jpg", + "4004321_01.jpg", + "4004321_02.jpg", + "4004321_03.jpg", + "4004321_04.jpg", + "4004321_05.jpg", + "4004321_06.jpg", + "4004321_07.jpg", + "4004321_08.jpg", + "4004321_09.jpg", + "4004321_10.jpg", + "4004321_11.jpg", + "4004321_12.jpg", + "4004321_13.jpg", + "4004321_14.jpg", + "4004321_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A person looking back at a closed door with a mixture of sadness and resolution. The door is illuminated by a soft, fading light, emphasizing the significance of the final glance before moving forward. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "3003219", + "images": [ + "3003219_00.jpg", + "3003219_01.jpg", + "3003219_02.jpg", + "3003219_03.jpg", + "3003219_04.jpg", + "3003219_05.jpg", + "3003219_06.jpg", + "3003219_07.jpg", + "3003219_08.jpg", + "3003219_09.jpg", + "3003219_10.jpg", + "3003219_11.jpg", + "3003219_12.jpg", + "3003219_13.jpg", + "3003219_14.jpg", + "3003219_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"elon musk standing ontop of a rainbow \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "3003801", + "images": [ + "3003801_00.jpg", + "3003801_01.jpg", + "3003801_02.jpg", + "3003801_03.jpg", + "3003801_04.jpg", + "3003801_05.jpg", + "3003801_06.jpg", + "3003801_07.jpg", + "3003801_08.jpg", + "3003801_09.jpg", + "3003801_10.jpg", + "3003801_11.jpg", + "3003801_12.jpg", + "3003801_13.jpg", + "3003801_14.jpg", + "3003801_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Earth revolving as we see from the space \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001404", + "images": [ + "8001404_00.jpg", + "8001404_01.jpg", + "8001404_02.jpg", + "8001404_03.jpg", + "8001404_04.jpg", + "8001404_05.jpg", + "8001404_06.jpg", + "8001404_07.jpg", + "8001404_08.jpg", + "8001404_09.jpg", + "8001404_10.jpg", + "8001404_11.jpg", + "8001404_12.jpg", + "8001404_13.jpg", + "8001404_14.jpg", + "8001404_15.jpg", + "8001404_16.jpg", + "8001404_17.jpg", + "8001404_18.jpg", + "8001404_19.jpg", + "8001404_20.jpg", + "8001404_21.jpg", + "8001404_22.jpg", + "8001404_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a very beautiful girl sitting in meditation \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1004509", + "images": [ + "1004509_00.jpg", + "1004509_01.jpg", + "1004509_02.jpg", + "1004509_03.jpg", + "1004509_04.jpg", + "1004509_05.jpg", + "1004509_06.jpg", + "1004509_07.jpg", + "1004509_08.jpg", + "1004509_09.jpg", + "1004509_10.jpg", + "1004509_11.jpg", + "1004509_12.jpg", + "1004509_13.jpg", + "1004509_14.jpg", + "1004509_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"kitten and kid both turning back and see \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "0003667", + "images": [ + "0003667_00.jpg", + "0003667_01.jpg", + "0003667_02.jpg", + "0003667_03.jpg", + "0003667_04.jpg", + "0003667_05.jpg", + "0003667_06.jpg", + "0003667_07.jpg", + "0003667_08.jpg", + "0003667_09.jpg", + "0003667_10.jpg", + "0003667_11.jpg", + "0003667_12.jpg", + "0003667_13.jpg", + "0003667_14.jpg", + "0003667_15.jpg", + "0003667_16.jpg", + "0003667_17.jpg", + "0003667_18.jpg", + "0003667_19.jpg", + "0003667_20.jpg", + "0003667_21.jpg", + "0003667_22.jpg", + "0003667_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a teenage girl wearing blue t shirt fixing phone in phone stand \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "3003218", + "images": [ + "3003218_00.jpg", + "3003218_01.jpg", + "3003218_02.jpg", + "3003218_03.jpg", + "3003218_04.jpg", + "3003218_05.jpg", + "3003218_06.jpg", + "3003218_07.jpg", + "3003218_08.jpg", + "3003218_09.jpg", + "3003218_10.jpg", + "3003218_11.jpg", + "3003218_12.jpg", + "3003218_13.jpg", + "3003218_14.jpg", + "3003218_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"imagine and create in motion the character of a cute tiny ukulele that has two eyes and a small mouth and that dances. In the background is a Mexican village at night with pretty festive mood lighting and pretty decorations. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "6003744", + "images": [ + "6003744_00.jpg", + "6003744_01.jpg", + "6003744_02.jpg", + "6003744_03.jpg", + "6003744_04.jpg", + "6003744_05.jpg", + "6003744_06.jpg", + "6003744_07.jpg", + "6003744_08.jpg", + "6003744_09.jpg", + "6003744_10.jpg", + "6003744_11.jpg", + "6003744_12.jpg", + "6003744_13.jpg", + "6003744_14.jpg", + "6003744_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"dog with wings fly in clouds sky \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001468", + "images": [ + "8001468_00.jpg", + "8001468_01.jpg", + "8001468_02.jpg", + "8001468_03.jpg", + "8001468_04.jpg", + "8001468_05.jpg", + "8001468_06.jpg", + "8001468_07.jpg", + "8001468_08.jpg", + "8001468_09.jpg", + "8001468_10.jpg", + "8001468_11.jpg", + "8001468_12.jpg", + "8001468_13.jpg", + "8001468_14.jpg", + "8001468_15.jpg", + "8001468_16.jpg", + "8001468_17.jpg", + "8001468_18.jpg", + "8001468_19.jpg", + "8001468_20.jpg", + "8001468_21.jpg", + "8001468_22.jpg", + "8001468_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"procurement and logistics company specialized in logistics for industrial goods to the regions of the Far North using various modes of transportation, including winter roads. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "7002366", + "images": [ + "7002366_00.jpg", + "7002366_01.jpg", + "7002366_02.jpg", + "7002366_03.jpg", + "7002366_04.jpg", + "7002366_05.jpg", + "7002366_06.jpg", + "7002366_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"\",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 3\n" + } + ] + }, + { + "id": "4003139", + "images": [ + "4003139_00.jpg", + "4003139_01.jpg", + "4003139_02.jpg", + "4003139_03.jpg", + "4003139_04.jpg", + "4003139_05.jpg", + "4003139_06.jpg", + "4003139_07.jpg", + "4003139_08.jpg", + "4003139_09.jpg", + "4003139_10.jpg", + "4003139_11.jpg", + "4003139_12.jpg", + "4003139_13.jpg", + "4003139_14.jpg", + "4003139_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a cat is lurking in the night \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "4003301", + "images": [ + "4003301_00.jpg", + "4003301_01.jpg", + "4003301_02.jpg", + "4003301_03.jpg", + "4003301_04.jpg", + "4003301_05.jpg", + "4003301_06.jpg", + "4003301_07.jpg", + "4003301_08.jpg", + "4003301_09.jpg", + "4003301_10.jpg", + "4003301_11.jpg", + "4003301_12.jpg", + "4003301_13.jpg", + "4003301_14.jpg", + "4003301_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a white dog and a black cat and yellow frog play together. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "4003317", + "images": [ + "4003317_00.jpg", + "4003317_01.jpg", + "4003317_02.jpg", + "4003317_03.jpg", + "4003317_04.jpg", + "4003317_05.jpg", + "4003317_06.jpg", + "4003317_07.jpg", + "4003317_08.jpg", + "4003317_09.jpg", + "4003317_10.jpg", + "4003317_11.jpg", + "4003317_12.jpg", + "4003317_13.jpg", + "4003317_14.jpg", + "4003317_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"castle , wall disney , wind \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "6003161", + "images": [ + "6003161_00.jpg", + "6003161_01.jpg", + "6003161_02.jpg", + "6003161_03.jpg", + "6003161_04.jpg", + "6003161_05.jpg", + "6003161_06.jpg", + "6003161_07.jpg", + "6003161_08.jpg", + "6003161_09.jpg", + "6003161_10.jpg", + "6003161_11.jpg", + "6003161_12.jpg", + "6003161_13.jpg", + "6003161_14.jpg", + "6003161_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The Devil drawn in the cartoon style of Don Hertzfeldt \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3003199", + "images": [ + "3003199_00.jpg", + "3003199_01.jpg", + "3003199_02.jpg", + "3003199_03.jpg", + "3003199_04.jpg", + "3003199_05.jpg", + "3003199_06.jpg", + "3003199_07.jpg", + "3003199_08.jpg", + "3003199_09.jpg", + "3003199_10.jpg", + "3003199_11.jpg", + "3003199_12.jpg", + "3003199_13.jpg", + "3003199_14.jpg", + "3003199_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"farmer digging for portal coins with a alien portal in the sky. use portal coin logo \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "6003329", + "images": [ + "6003329_00.jpg", + "6003329_01.jpg", + "6003329_02.jpg", + "6003329_03.jpg", + "6003329_04.jpg", + "6003329_05.jpg", + "6003329_06.jpg", + "6003329_07.jpg", + "6003329_08.jpg", + "6003329_09.jpg", + "6003329_10.jpg", + "6003329_11.jpg", + "6003329_12.jpg", + "6003329_13.jpg", + "6003329_14.jpg", + "6003329_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Mario eating spaghetti for 4 hours \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "3005701", + "images": [ + "3005701_00.jpg", + "3005701_01.jpg", + "3005701_02.jpg", + "3005701_03.jpg", + "3005701_04.jpg", + "3005701_05.jpg", + "3005701_06.jpg", + "3005701_07.jpg", + "3005701_08.jpg", + "3005701_09.jpg", + "3005701_10.jpg", + "3005701_11.jpg", + "3005701_12.jpg", + "3005701_13.jpg", + "3005701_14.jpg", + "3005701_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"gatekeeper denied entry to four people \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 1\n text-to-video alignment: 1\n factual consistency: 2\n" + } + ] + }, + { + "id": "6004531", + "images": [ + "6004531_00.jpg", + "6004531_01.jpg", + "6004531_02.jpg", + "6004531_03.jpg", + "6004531_04.jpg", + "6004531_05.jpg", + "6004531_06.jpg", + "6004531_07.jpg", + "6004531_08.jpg", + "6004531_09.jpg", + "6004531_10.jpg", + "6004531_11.jpg", + "6004531_12.jpg", + "6004531_13.jpg", + "6004531_14.jpg", + "6004531_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"With a heavy heart, the king realized the depth of his folly(3D cartoon video \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "7004046", + "images": [ + "7004046_00.jpg", + "7004046_01.jpg", + "7004046_02.jpg", + "7004046_03.jpg", + "7004046_04.jpg", + "7004046_05.jpg", + "7004046_06.jpg", + "7004046_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"As the wine swayed, the picture began to blur, and the color gradually transitioned from the crimson of the wine to the light color of the pavilion. Design a gradient effect between the wine and the pavilion lights. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 3\n" + } + ] + }, + { + "id": "7004328", + "images": [ + "7004328_00.jpg", + "7004328_01.jpg", + "7004328_02.jpg", + "7004328_03.jpg", + "7004328_04.jpg", + "7004328_05.jpg", + "7004328_06.jpg", + "7004328_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A beautiful Chinese beauty, non animated \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + }, + { + "id": "1006943", + "images": [ + "1006943_00.jpg", + "1006943_01.jpg", + "1006943_02.jpg", + "1006943_03.jpg", + "1006943_04.jpg", + "1006943_05.jpg", + "1006943_06.jpg", + "1006943_07.jpg", + "1006943_08.jpg", + "1006943_09.jpg", + "1006943_10.jpg", + "1006943_11.jpg", + "1006943_12.jpg", + "1006943_13.jpg", + "1006943_14.jpg", + "1006943_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"An Asian girl is drinking beer while visiting the night market at night \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "8001625", + "images": [ + "8001625_00.jpg", + "8001625_01.jpg", + "8001625_02.jpg", + "8001625_03.jpg", + "8001625_04.jpg", + "8001625_05.jpg", + "8001625_06.jpg", + "8001625_07.jpg", + "8001625_08.jpg", + "8001625_09.jpg", + "8001625_10.jpg", + "8001625_11.jpg", + "8001625_12.jpg", + "8001625_13.jpg", + "8001625_14.jpg", + "8001625_15.jpg", + "8001625_16.jpg", + "8001625_17.jpg", + "8001625_18.jpg", + "8001625_19.jpg", + "8001625_20.jpg", + "8001625_21.jpg", + "8001625_22.jpg", + "8001625_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Many demons start emerge out of the burning demonic circle, night, cinematic, birds eye perspective \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 2\n factual consistency: 1\n" + } + ] + }, + { + "id": "5001522", + "images": [ + "5001522_00.jpg", + "5001522_01.jpg", + "5001522_02.jpg", + "5001522_03.jpg", + "5001522_04.jpg", + "5001522_05.jpg", + "5001522_06.jpg", + "5001522_07.jpg", + "5001522_08.jpg", + "5001522_09.jpg", + "5001522_10.jpg", + "5001522_11.jpg", + "5001522_12.jpg", + "5001522_13.jpg", + "5001522_14.jpg", + "5001522_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Mafia boss Entry in hotel apartment, aerial view, cinematic shot, London Street, 64k \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 2\n" + } + ] + }, + { + "id": "8002946", + "images": [ + "8002946_00.jpg", + "8002946_01.jpg", + "8002946_02.jpg", + "8002946_03.jpg", + "8002946_04.jpg", + "8002946_05.jpg", + "8002946_06.jpg", + "8002946_07.jpg", + "8002946_08.jpg", + "8002946_09.jpg", + "8002946_10.jpg", + "8002946_11.jpg", + "8002946_12.jpg", + "8002946_13.jpg", + "8002946_14.jpg", + "8002946_15.jpg", + "8002946_16.jpg", + "8002946_17.jpg", + "8002946_18.jpg", + "8002946_19.jpg", + "8002946_20.jpg", + "8002946_21.jpg", + "8002946_22.jpg", + "8002946_23.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a video of Sasuke Uchiha making a Chidori, 4k, ultrifine detail. Cinematic lighting \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n" + } + ] + }, + { + "id": "3003847", + "images": [ + "3003847_00.jpg", + "3003847_01.jpg", + "3003847_02.jpg", + "3003847_03.jpg", + "3003847_04.jpg", + "3003847_05.jpg", + "3003847_06.jpg", + "3003847_07.jpg", + "3003847_08.jpg", + "3003847_09.jpg", + "3003847_10.jpg", + "3003847_11.jpg", + "3003847_12.jpg", + "3003847_13.jpg", + "3003847_14.jpg", + "3003847_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The bank lobby is bustling with customers. Security guards patrol the area. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "1006142", + "images": [ + "1006142_00.jpg", + "1006142_01.jpg", + "1006142_02.jpg", + "1006142_03.jpg", + "1006142_04.jpg", + "1006142_05.jpg", + "1006142_06.jpg", + "1006142_07.jpg", + "1006142_08.jpg", + "1006142_09.jpg", + "1006142_10.jpg", + "1006142_11.jpg", + "1006142_12.jpg", + "1006142_13.jpg", + "1006142_14.jpg", + "1006142_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Show a charming little girl, wearing a colorful ski outfit, gracefully skiing down the slopes, enjoying the snow, and an energetic little boy, with his cool snowboarding gear, pulling off exciting snowboarding tricks. Capture their expressions of pure joy and enthusiasm. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n" + } + ] + }, + { + "id": "7004531", + "images": [ + "7004531_00.jpg", + "7004531_01.jpg", + "7004531_02.jpg", + "7004531_03.jpg", + "7004531_04.jpg", + "7004531_05.jpg", + "7004531_06.jpg", + "7004531_07.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"and from between the trees emerged the fearsome silhouette of a Tyrannosaurus rex. Its massive jaws moving and dripped with saliva, and its sharp teeth glistened in the fading light. \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n" + } + ] + }, + { + "id": "2005387", + "images": [ + "2005387_00.jpg", + "2005387_01.jpg", + "2005387_02.jpg", + "2005387_03.jpg", + "2005387_04.jpg", + "2005387_05.jpg", + "2005387_06.jpg", + "2005387_07.jpg", + "2005387_08.jpg", + "2005387_09.jpg", + "2005387_10.jpg", + "2005387_11.jpg", + "2005387_12.jpg", + "2005387_13.jpg", + "2005387_14.jpg", + "2005387_15.jpg" + ], + "conversations": [ + { + "from": "human", + "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Timelapse of rotating stars in the Sky, HDR, cinematic frame \",\nall the frames of video are as follows: \n\n \n" + }, + { + "from": "gpt", + "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n" + } + ] + } +] \ No newline at end of file