[
 {
  "video_id": "insave_general_editing_48600",
  "dataset": "InsAVE-80K",
  "dataset_split": "general_editing",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_general_editing_48600.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/general_editing/part_04.tar::48600_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Sheep",
  "duration_s": 5.043,
  "audio_rms": 0.006568,
  "sound_tags": "Sheep:0.6727;Bleat:0.5872;Animal:0.5728;Goat:0.5506;Livestock, farm animals, working animals:0.1982",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 24.0,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_eval_00231",
  "dataset": "InsAVE-80K",
  "dataset_split": "eval",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_eval_00231.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/eval_data/eval/part_00.tar::00231_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Animal",
  "duration_s": 3.397,
  "audio_rms": 0.016131,
  "sound_tags": "Animal:0.1191;Speech:0.114;Door:0.1097;Bird:0.0712;Chicken, rooster:0.0577",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_add_and_remove_04959",
  "dataset": "InsAVE-80K",
  "dataset_split": "add_and_remove",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_add_and_remove_04959.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/add_and_remove/part_00.tar::04959_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Speech",
  "duration_s": 3.397,
  "audio_rms": 0.006925,
  "sound_tags": "Speech:0.3696;Animal:0.0651;Sneeze:0.0648;Gasp:0.0355;Clip-clop:0.0314",
  "no_speech": {
   "vad_speech_seconds": 0.96,
   "asr_checked": true,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 24.0,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "javedit_7162ff62de8f",
  "dataset": "JAVEdit-100k",
  "dataset_split": "source",
  "video": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/7162ff62de8f4f6951087cf03ce53004.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/7162ff62de8f4f6951087cf03ce53004.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Speech",
  "duration_s": 4.864,
  "audio_rms": 0.004911,
  "sound_tags": "Speech:0.8139;Silence:0.0765;Television:0.0577;Male speech, man speaking:0.0529;Inside, small room:0.0333",
  "no_speech": {
   "vad_speech_seconds": 1.23,
   "asr_checked": true,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 720,
   "fps": 25.0,
   "audio": "aac 16kHz mono"
  }
 },
 {
  "video_id": "insave_general_editing_34794",
  "dataset": "InsAVE-80K",
  "dataset_split": "general_editing",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_general_editing_34794.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/general_editing/part_02.tar::34794_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Speech",
  "duration_s": 6.715,
  "audio_rms": 0.009234,
  "sound_tags": "Speech:0.6605;Animal:0.2845;Bird:0.0996;Duck:0.0536;Quack:0.0457",
  "no_speech": {
   "vad_speech_seconds": 2.18,
   "asr_checked": true,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_add_and_remove_07472",
  "dataset": "InsAVE-80K",
  "dataset_split": "add_and_remove",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_add_and_remove_07472.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/add_and_remove/part_00.tar::07472_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Door",
  "duration_s": 5.069,
  "audio_rms": 0.020616,
  "sound_tags": "Door:0.2988;Slam:0.1661;Walk, footsteps:0.0877;Thunk:0.0799;Sliding door:0.0549",
  "no_speech": {
   "vad_speech_seconds": 0.61,
   "asr_checked": true,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_general_editing_31165",
  "dataset": "InsAVE-80K",
  "dataset_split": "general_editing",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_general_editing_31165.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/general_editing/part_02.tar::31165_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Liquid",
  "duration_s": 3.397,
  "audio_rms": 0.002911,
  "sound_tags": "Liquid:0.1461;Rowboat, canoe, kayak:0.1105;Boat, Water vehicle:0.0855;Stream:0.0803;Animal:0.0772",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_general_editing_47918",
  "dataset": "InsAVE-80K",
  "dataset_split": "general_editing",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_general_editing_47918.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/general_editing/part_04.tar::47918_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Frog",
  "duration_s": 3.397,
  "audio_rms": 0.001884,
  "sound_tags": "Frog:0.2983;Croak:0.107;Animal:0.0503;Walk, footsteps:0.0439;Outside, rural or natural:0.0439",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "insave_add_and_remove_04160",
  "dataset": "InsAVE-80K",
  "dataset_split": "add_and_remove",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_add_and_remove_04160.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/add_and_remove/part_00.tar::04160_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Typewriter",
  "duration_s": 6.715,
  "audio_rms": 0.001808,
  "sound_tags": "Typewriter:0.2837;Door:0.2652;Sliding door:0.119;Printer:0.0366;Cash register:0.0364",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 },
 {
  "video_id": "javedit_5b6a0a248356",
  "dataset": "JAVEdit-100k",
  "dataset_split": "source",
  "video": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/5b6a0a248356a61f89239578f68d08a8.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/5b6a0a248356a61f89239578f68d08a8.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Clip-clop",
  "duration_s": 4.864,
  "audio_rms": 0.002357,
  "sound_tags": "Clip-clop:0.626;Animal:0.5685;Horse:0.4016;Walk, footsteps:0.1041;Outside, rural or natural:0.0775",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 720,
   "fps": 25.0,
   "audio": "aac 16kHz mono"
  }
 },
 {
  "video_id": "javedit_4830e7822b49",
  "dataset": "JAVEdit-100k",
  "dataset_split": "source",
  "video": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/4830e7822b497f48940be6a934998afc.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/JAVEdit-100k/source_videos/4830e7822b497f48940be6a934998afc.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Vehicle",
  "duration_s": 4.864,
  "audio_rms": 0.054473,
  "sound_tags": "Vehicle:0.5236;Car:0.131;Truck:0.065;Medium engine (mid frequency):0.056;Bus:0.0542",
  "no_speech": {
   "vad_speech_seconds": 0.0,
   "asr_checked": false,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 720,
   "fps": 25.0,
   "audio": "aac 16kHz mono"
  }
 },
 {
  "video_id": "insave_add_and_remove_05404",
  "dataset": "InsAVE-80K",
  "dataset_split": "add_and_remove",
  "video": "/group2/ct/weihanx/av_langgraph_pipeline/data/raw/insave/insave_add_and_remove_05404.mp4",
  "source": "/group2/ct/weihanx/baseline_dataset/InsAVE-80K/train_data/add_and_remove/part_00.tar::05404_original.mp4",
  "audio_src": null,
  "video_src": null,
  "audio_tgt": null,
  "video_tgt": null,
  "text": "Animal",
  "duration_s": 3.397,
  "audio_rms": 0.033219,
  "sound_tags": "Animal:0.3916;Domestic animals, pets:0.2741;Bow-wow:0.2555;Dog:0.2078;Speech:0.1509",
  "no_speech": {
   "vad_speech_seconds": 0.53,
   "asr_checked": true,
   "asr_text": ""
  },
  "music_score": 0.0,
  "meta": {
   "width": 1280,
   "height": 704,
   "fps": 23.98,
   "audio": "mp3 44kHz stereo"
  }
 }
]