Download inputs/shot_09/input.json from ymyy307/testltx: direct link, hf CLI and curl.
- Browser
- Download file 8.19 kB
-
https://huggingface.co/ymyy307/testltx/resolve/main/inputs/shot_09/input.json
- Command line
-
hf download hf://ymyy307/testltx/inputs/shot_09/input.json
-
curl -L -o input.json https://huggingface.co/ymyy307/testltx/resolve/main/inputs/shot_09/input.json
8.19 kB
| { | |
| "shot_index": 9, | |
| "raw_prompt": "<Subject 2> is an older East Asian male birdwatcher with neatly cut gray hair, a brown waterproof coat, and compact binoculars. <Subject 3> is a young white female audio student with auburn hair in a low ponytail, a navy rain shell, and small over-ear field headphones. At normal speed, they take the main branch together, walking slowly past water visible between the boards. <Subject 3> keeps one headphone cup off her ear. No intelligible dialogue is heard. The shot uses realistic cinematic imagery, physically plausible restrained motion, and a natural documentary-like tone. The audible soundscape is: Alternating footsteps, water below, and reed hiss.", | |
| "selector_query": "<Subject 1> is an older East Asian male birdwatcher with neatly cut gray hair, a brown waterproof coat, and compact binoculars. <Subject 2> is a young white female audio student with auburn hair in a low ponytail, a navy rain shell, and small over-ear field headphones. At normal speed, they take the main branch together, walking slowly past water visible between the boards. <Subject 2> keeps one headphone cup off her ear. No intelligible dialogue is heard. The shot uses realistic cinematic imagery, physically plausible restrained motion, and a natural documentary-like tone. The audible soundscape is: Alternating footsteps, water below, and reed hiss. Identity continuity requirements: <Subject 1> matches his fixed reference exactly: an elderly East Asian man around seventy with light-medium skin, a broad oval-to-square face, narrow dark-brown eyes, visible upper-lid folds and under-eye bags, a low broad nose bridge with a rounded tip, medium-thin lips, a softly squared jaw, natural age lines, and short silver-gray hair combed neatly to his left. He wears the same brown waterproof coat and carries compact binoculars. <Subject 2> matches her fixed reference and the correct appearance established in shot 7: a young European white woman in her late teens or early twenties with pale cool-toned skin, a softly oval face, naturally full cheeks, softly set light gray-green eyes, a straight medium-width nose with a rounded tip, small natural pink lips, and a small rounded chin. Her copper-auburn hair has a clean center part and is smoothly swept away from her fully exposed forehead and both temples into a low ponytail. She wears the same navy rain shell and small black over-ear field headphones. Fixed reference-image mapping: Picture 1 exclusively fixes <Subject 1>. Picture 2 exclusively fixes <Subject 2> from the tightly cropped locked shot-7 three-quarter reference. Preserve exactly two people and transfer the complete facial identity from each corresponding Picture. Composition requirement: Film one continuous full-frame live-action medium tracking shot. In the first frame, <Subject 1> is already fully visible on frame-left and <Subject 2> is already fully visible on frame-right, walking side by side at equal distance from the camera. The camera tracks backward at exactly their pace. Keep both people centered with generous outer margins, fully visible from head to waist, spatially separated, equally sized, and in the same left-right positions through the final frame. Preserve the elderly man's exact face, silver-gray hair, brown coat, and binoculars. Preserve the exact shot-7 girl's face, smooth exposed forehead, copper-auburn low ponytail, navy rain shell, and complete black headphones; one full-size ear cup rests slightly away from her ear while the headband and both cups remain connected. The location, lighting, and exposure remain naturally continuous throughout.", | |
| "generation_prompt": "<Subject 1> is an older East Asian male birdwatcher with neatly cut gray hair, a brown waterproof coat, and compact binoculars. <Subject 2> is a young white female audio student with auburn hair in a low ponytail, a navy rain shell, and small over-ear field headphones. At normal speed, they take the main branch together, walking slowly past water visible between the boards. <Subject 2> keeps one headphone cup off her ear. No intelligible dialogue is heard. The shot uses realistic cinematic imagery, physically plausible restrained motion, and a natural documentary-like tone. The audible soundscape is: Alternating footsteps, water below, and reed hiss. Identity continuity requirements: <Subject 1> matches his fixed reference exactly: an elderly East Asian man around seventy with light-medium skin, a broad oval-to-square face, narrow dark-brown eyes, visible upper-lid folds and under-eye bags, a low broad nose bridge with a rounded tip, medium-thin lips, a softly squared jaw, natural age lines, and short silver-gray hair combed neatly to his left. He wears the same brown waterproof coat and carries compact binoculars. <Subject 2> matches her fixed reference and the correct appearance established in shot 7: a young European white woman in her late teens or early twenties with pale cool-toned skin, a softly oval face, naturally full cheeks, softly set light gray-green eyes, a straight medium-width nose with a rounded tip, small natural pink lips, and a small rounded chin. Her copper-auburn hair has a clean center part and is smoothly swept away from her fully exposed forehead and both temples into a low ponytail. She wears the same navy rain shell and small black over-ear field headphones. Fixed reference-image mapping: Picture 1 exclusively fixes <Subject 1>. Picture 2 exclusively fixes <Subject 2> from the tightly cropped locked shot-7 three-quarter reference. Preserve exactly two people and transfer the complete facial identity from each corresponding Picture. Composition requirement: Film one continuous full-frame live-action medium tracking shot. In the first frame, <Subject 1> is already fully visible on frame-left and <Subject 2> is already fully visible on frame-right, walking side by side at equal distance from the camera. The camera tracks backward at exactly their pace. Keep both people centered with generous outer margins, fully visible from head to waist, spatially separated, equally sized, and in the same left-right positions through the final frame. Preserve the elderly man's exact face, silver-gray hair, brown coat, and binoculars. Preserve the exact shot-7 girl's face, smooth exposed forehead, copper-auburn low ponytail, navy rain shell, and complete black headphones; one full-size ear cup rests slightly away from her ear while the headband and both cups remain connected. The location, lighting, and exposure remain naturally continuous throughout.", | |
| "pictures": [ | |
| { | |
| "picture": 1, | |
| "path": "rerun_identity_anchor_01.png", | |
| "source_shot": null, | |
| "frame_index": null, | |
| "identity_anchor": "<Subject 1>", | |
| "source": "/datadrive1/v-zihanyang/evals/unityshotsbench/ltx23_joyai_echo_subject_labels_20261111/refine_assets/worker_04_tier3/subject_1_man_close_shot02_frame0060.png" | |
| }, | |
| { | |
| "picture": 2, | |
| "path": "rerun_identity_anchor_02.png", | |
| "source_shot": null, | |
| "frame_index": null, | |
| "identity_anchor": "<Subject 2 Three Quarter>", | |
| "source": "/datadrive1/v-zihanyang/evals/unityshotsbench/ltx23_joyai_echo_subject_labels_20261111/refine_assets/worker_04_tier3/subject_2_girl_fixed_shot07_frame000_headshoulders.png" | |
| } | |
| ], | |
| "audio_references": [ | |
| { | |
| "label": "<Audio 1>", | |
| "source_shot_idx": 8, | |
| "duration_sec": 5.041666666666667, | |
| "input_path": "audio_01_from_shot_08.wav" | |
| } | |
| ], | |
| "generation": { | |
| "model": "/datadrive1/v-zihanyang/Audio-Video-Benchmarks/LTX-2/models/ltx-2.3-22b-dev.safetensors", | |
| "lora": "/datadrive2/v-zihanyang/restored_training_local/ltx23/rl_checkpoint61_peft/lora_bf16.safetensors", | |
| "width": 1280, | |
| "height": 736, | |
| "num_frames": 121, | |
| "fps": 24, | |
| "num_inference_steps": 30, | |
| "seed": 6809, | |
| "video_cfg_scale": 4.0, | |
| "audio_cfg_scale": 7.0, | |
| "video_stg_scale": 1.0, | |
| "audio_stg_scale": 1.0, | |
| "stg_blocks": [ | |
| 29 | |
| ], | |
| "guidance_rescale": 0.7, | |
| "video_modality_guidance_scale": 1.0, | |
| "audio_modality_guidance_scale": 1.0, | |
| "token_order": "prepend" | |
| }, | |
| "anchor_only_refinement": true | |
| } | |