record,json_value
test_plan,"{""sequence"": [""D23-1"", ""D23-2"", ""D23-3"", ""D24-1"", ""D24-2"", ""D24-3""], ""test_date"": ""2026-10-15"", ""hardware"": ""NVIDIA GeForce RTX 3090 24576 MiB; AMD EPYC 7402P; driver 595.84; power limit 275 W"", ""latent"": {""release"": ""1.6"", ""commit"": ""a229c3948406bc2cf6eaf4873e662e70c6a04746"", ""model_revision"": ""c42c7e6c8e9c213626389fa7d9a3c444b8536353"", ""steps"": 20, ""guidance_scale"": 1.5, ""seed"": 1247, ""deepcache"": true, ""face_resolution"": 512}, ""muse"": {""release"": ""1.5"", ""commit"": ""0a89dec45a0192b824e3cf4daf96c239440c5ed8"", ""model_revision"": ""2bcb936e2fddb4d86db4c62fd45b387d0c061571"", ""batch_size"": 2, ""precision"": ""fp16"", ""bbox_shift"": 0, ""extra_margin"": 10, ""parsing_mode"": ""jaw"", ""left_cheek_width"": 90, ""right_cheek_width"": 90, ""face_resolution"": 256}, ""sample"": ""Shared D18 permitted fictional stationary-face input.mp4 and synthetic speech.wav, identical original bytes for both models. One underlying scene and utterance; no claim of dataset breadth."", ""schedule"": [""latent-initial"", ""muse-initial"", ""muse-r2"", ""latent-r2"", ""latent-r3"", ""muse-r3""], ""timing"": ""Fresh process start through exit including imports, checkpoint loading, preprocessing, generation and encoding. Excludes environment/model provisioning, transfer and independent validation. Later runs have warm OS/model caches; initial run may fetch auxiliary dependencies, recorded separately if encountered."", ""fairness_limits"": [""Different native face resolutions, model architectures, dependencies and inference controls"", ""No equal compute budget or control sweep"", ""Idle pre-existing ComfyUI process kept running; baseline VRAM and activity recorded"", ""Single static source does not assess head motion, profiles, languages or identities"", ""No independent perceptual or sync metric; visual judgments labeled subjective""], ""screenshots"": ""Real browser captures of recorded files and paired actual frames, original screenshots retained with separate arrow copies. Arrows only in reserved clear margin.""}"
setup_exceptions,"{""decord_metadata"": {""condition"": ""decord 0.6.0 official py3 wheel embeds cp36-cp36m-manylinux2010_x86_64"", ""pip_check"": ""flags unsupported platform"", ""uv_check"": ""flags same platform issue"", ""dependency_constraints"": ""pip check_package_set returned ({}, {}): no missing or conflicting requirements"", ""runtime"": ""91 frames decoded at 512x512 under Python 3.10.13; stock inference imports and CUDA tensor succeeded"", ""change"": ""none to wheel or model source""}, ""onnx_cuda"": {""initial_condition"": ""libnvrtc.so.12 not on dynamic loader path; initial preparation fell back to CPU"", ""correction"": ""LD_LIBRARY_PATH explicitly includes installed latent-env site-packages/nvidia/*/lib"", ""verified"": ""detection and landmark_2d_106 sessions report CUDAExecutionProvider first; real sample face detected before timing"", ""source_changes"": ""none""}}"
input_manifest,"{""comparison.yaml"": {""bytes"": 155, ""sha256"": ""513ac7318b3e76c56aa91b02501810296284435a1ccc516b5ab5e95029057045""}, ""input.mp4"": {""bytes"": 44591, ""sha256"": ""895c94e8efff634b054abd17951e0c6e9cb8c1e864a0a9e4035b832f4a676a26""}, ""portrait.png"": {""bytes"": 2039351, ""sha256"": ""029185678e4435ee0501ebff4c40ed4753b2dbfba6fc42e16e906510428e1767""}, ""speech.wav"": {""bytes"": 116296, ""sha256"": ""89a902c22a5d0d75e8c37b1359ed5f71221d16f89fc9ca8e84b6ee1b336bec62""}, ""transcript.txt"": {""bytes"": 51, ""sha256"": ""b73f23ba36237d813bab949c827015cbb1af4143a24d230e7abc29848d1a6b91""}}"
commands,"{""latent-r1"": [""/workspace/d23-runtime/latent-env/bin/python"", ""-m"", ""scripts.inference"", ""--unet_config_path"", ""configs/unet/stage2_512.yaml"", ""--inference_ckpt_path"", ""checkpoints/latentsync_unet.pt"", ""--inference_steps"", ""20"", ""--guidance_scale"", ""1.5"", ""--seed"", ""1247"", ""--enable_deepcache"", ""--video_path"", ""/workspace/d23-kit/inputs/input.mp4"", ""--audio_path"", ""/workspace/d23-kit/inputs/speech.wav"", ""--video_out_path"", ""/workspace/d23-evidence/latent-r1/result.mp4"", ""--temp_dir"", ""/workspace/d23-evidence/latent-r1/temp""], ""muse-r1"": [""/workspace/d23-runtime/muse-env/bin/python"", ""-m"", ""scripts.inference"", ""--inference_config"", ""/workspace/d23-kit/inputs/comparison.yaml"", ""--result_dir"", ""/workspace/d23-evidence/muse-r1/generated"", ""--unet_model_path"", ""models/musetalkV15/unet.pth"", ""--unet_config"", ""models/musetalkV15/musetalk.json"", ""--whisper_dir"", ""models/whisper"", ""--version"", ""v15"", ""--batch_size"", ""2"", ""--use_float16"", ""--bbox_shift"", ""0"", ""--fps"", ""25"", ""--extra_margin"", ""10"", ""--parsing_mode"", ""jaw"", ""--left_cheek_width"", ""90"", ""--right_cheek_width"", ""90"", ""--ffmpeg_path"", ""/usr/bin""], ""muse-r2"": [""/workspace/d23-runtime/muse-env/bin/python"", ""-m"", ""scripts.inference"", ""--inference_config"", ""/workspace/d23-kit/inputs/comparison.yaml"", ""--result_dir"", ""/workspace/d23-evidence/muse-r2/generated"", ""--unet_model_path"", ""models/musetalkV15/unet.pth"", ""--unet_config"", ""models/musetalkV15/musetalk.json"", ""--whisper_dir"", ""models/whisper"", ""--version"", ""v15"", ""--batch_size"", ""2"", ""--use_float16"", ""--bbox_shift"", ""0"", ""--fps"", ""25"", ""--extra_margin"", ""10"", ""--parsing_mode"", ""jaw"", ""--left_cheek_width"", ""90"", ""--right_cheek_width"", ""90"", ""--ffmpeg_path"", ""/usr/bin""], ""latent-r2"": [""/workspace/d23-runtime/latent-env/bin/python"", ""-m"", ""scripts.inference"", ""--unet_config_path"", ""configs/unet/stage2_512.yaml"", ""--inference_ckpt_path"", ""checkpoints/latentsync_unet.pt"", ""--inference_steps"", ""20"", ""--guidance_scale"", ""1.5"", ""--seed"", ""1247"", ""--enable_deepcache"", ""--video_path"", ""/workspace/d23-kit/inputs/input.mp4"", ""--audio_path"", ""/workspace/d23-kit/inputs/speech.wav"", ""--video_out_path"", ""/workspace/d23-evidence/latent-r2/result.mp4"", ""--temp_dir"", ""/workspace/d23-evidence/latent-r2/temp""], ""latent-r3"": [""/workspace/d23-runtime/latent-env/bin/python"", ""-m"", ""scripts.inference"", ""--unet_config_path"", ""configs/unet/stage2_512.yaml"", ""--inference_ckpt_path"", ""checkpoints/latentsync_unet.pt"", ""--inference_steps"", ""20"", ""--guidance_scale"", ""1.5"", ""--seed"", ""1247"", ""--enable_deepcache"", ""--video_path"", ""/workspace/d23-kit/inputs/input.mp4"", ""--audio_path"", ""/workspace/d23-kit/inputs/speech.wav"", ""--video_out_path"", ""/workspace/d23-evidence/latent-r3/result.mp4"", ""--temp_dir"", ""/workspace/d23-evidence/latent-r3/temp""], ""muse-r3"": [""/workspace/d23-runtime/muse-env/bin/python"", ""-m"", ""scripts.inference"", ""--inference_config"", ""/workspace/d23-kit/inputs/comparison.yaml"", ""--result_dir"", ""/workspace/d23-evidence/muse-r3/generated"", ""--unet_model_path"", ""models/musetalkV15/unet.pth"", ""--unet_config"", ""models/musetalkV15/musetalk.json"", ""--whisper_dir"", ""models/whisper"", ""--version"", ""v15"", ""--batch_size"", ""2"", ""--use_float16"", ""--bbox_shift"", ""0"", ""--fps"", ""25"", ""--extra_margin"", ""10"", ""--parsing_mode"", ""jaw"", ""--left_cheek_width"", ""90"", ""--right_cheek_width"", ""90"", ""--ffmpeg_path"", ""/usr/bin""]}"
latent_model_manifest,"[{""repo"": ""ByteDance/LatentSync-1.6"", ""revision"": ""c42c7e6c8e9c213626389fa7d9a3c444b8536353"", ""filename"": ""latentsync_unet.pt"", ""destination"": ""checkpoints/latentsync_unet.pt"", ""size"": 5072222488, ""sha256"": ""0a478e89eb660f82da4c35dbdde8a5adfb27f99d1b4e50edd03729e1e98316d3""}, {""repo"": ""ByteDance/LatentSync-1.6"", ""revision"": ""c42c7e6c8e9c213626389fa7d9a3c444b8536353"", ""filename"": ""whisper/tiny.pt"", ""destination"": ""checkpoints/whisper/tiny.pt"", ""size"": 75572083, ""sha256"": ""65147644a518d12f04e32d6f3b26facc3f8dd46e5390956a9424a650c0ce22b9""}]"
muse_model_manifest,"[{""repo"": ""TMElyralab/MuseTalk"", ""revision"": ""2bcb936e2fddb4d86db4c62fd45b387d0c061571"", ""filename"": ""musetalkV15/musetalk.json"", ""destination"": ""models/musetalkV15/musetalk.json"", ""size"": 748, ""sha256"": ""5b6923aee04d71692e0e9846c471e0a4ea07a4f686d39545e472bd4ba17e1b47""}, {""repo"": ""TMElyralab/MuseTalk"", ""revision"": ""2bcb936e2fddb4d86db4c62fd45b387d0c061571"", ""filename"": ""musetalkV15/unet.pth"", ""destination"": ""models/musetalkV15/unet.pth"", ""size"": 3400074924, ""sha256"": ""7ebf6c98c181e20838e4c0054e96e944ac60d5d692cc01db42839fe11b787007""}, {""repo"": ""stabilityai/sd-vae-ft-mse"", ""revision"": ""31f26fdeee1355a5c34592e401dd41e45d25a493"", ""filename"": ""config.json"", ""destination"": ""models/sd-vae/config.json"", ""size"": 547, ""sha256"": ""92d3dfb746fca211a2c9e019e285f8597412211728dce3c5bcf4eda0f2d62e7e""}, {""repo"": ""stabilityai/sd-vae-ft-mse"", ""revision"": ""31f26fdeee1355a5c34592e401dd41e45d25a493"", ""filename"": ""diffusion_pytorch_model.bin"", ""destination"": ""models/sd-vae/diffusion_pytorch_model.bin"", ""size"": 334707217, ""sha256"": ""1b4889b6b1d4ce7ae320a02dedaeff1780ad77d415ea0d744b476155c6377ddc""}, {""repo"": ""openai/whisper-tiny"", ""revision"": ""169d4a4341b33bc18d8881c4b69c2e104e1cc0af"", ""filename"": ""config.json"", ""destination"": ""models/whisper/config.json"", ""size"": 1983, ""sha256"": ""ffdccec4f3211f4c63310f2b7098f309fe70f3952cedc5e4d11e43f5b2379b98""}, {""repo"": ""openai/whisper-tiny"", ""revision"": ""169d4a4341b33bc18d8881c4b69c2e104e1cc0af"", ""filename"": ""preprocessor_config.json"", ""destination"": ""models/whisper/preprocessor_config.json"", ""size"": 184990, ""sha256"": ""9b5cd03a36fbb8a627c64d98a5b5b126ead95a77720723944487311f0110b666""}, {""repo"": ""openai/whisper-tiny"", ""revision"": ""169d4a4341b33bc18d8881c4b69c2e104e1cc0af"", ""filename"": ""pytorch_model.bin"", ""destination"": ""models/whisper/pytorch_model.bin"", ""size"": 151095027, ""sha256"": ""9607f98a2b22d9e229ae43c52ecea79dcede9e0c5cfae67e8da6eda86d8aac1d""}, {""repo"": ""yzd-v/DWPose"", ""revision"": ""1a7144101628d69ee7a3768d1ee3a094070dc388"", ""filename"": ""dw-ll_ucoco_384.pth"", ""destination"": ""models/dwpose/dw-ll_ucoco_384.pth"", ""size"": 406878486, ""sha256"": ""0d9408b13cd863c4e95a149dd31232f88f2a12aa6cf8964ed74d7d97748c7a07""}, {""repo"": ""ManyOtherFunctions/face-parse-bisent"", ""revision"": ""0073b233a5a3c4b1377d4dbf49245017938a72b5"", ""filename"": ""79999_iter.pth"", ""destination"": ""models/face-parse-bisent/79999_iter.pth"", ""size"": 53289463, ""sha256"": ""468e13ca13a9b43cc0881a9f99083a430e9c0a38abd935431d1c28ee94b26567""}, {""repo"": ""ManyOtherFunctions/face-parse-bisent"", ""revision"": ""0073b233a5a3c4b1377d4dbf49245017938a72b5"", ""filename"": ""resnet18-5c106cde.pth"", ""destination"": ""models/face-parse-bisent/resnet18-5c106cde.pth"", ""size"": 46827520, ""sha256"": ""5c106cde386e87d4033832f2996f5493238eda96ccf559d1d62760c4de0613f8""}]"