Portable desktop UI for Wav2Lip workflow (Windows).
Note: This is a fully portable solution. For GPU mode, installed NVIDIA/CUDA drivers are required on the target system.
| Main UI | Jobs / Pipeline |
|---|---|
![]() |
![]() |
- Upload source video and audio.
- Probe frame and select target face.
- Build target-face tracking.
- Save timeline segments.
- Create and run jobs (local queue worker).
- Supports extended input containers.
- Auto-normalizes problematic video codecs with FFmpeg fallback.
- CPU:
only_cpu.bat - GPU:
only_gpu.bat - Debug CPU:
debug_cpu.bat - Debug GPU:
debug_gpu.bat
portable_check.batThis GitHub repository is source-only. Heavy portable artifacts are intentionally excluded from git:
app/python(embedded Python runtime)app/bin/ffmpeg(bundled FFmpeg/FFprobe)Wav2Lip/checkpoints(model weights)- runtime outputs/caches (
app/data/*,app/logs)
Distribute full portable build separately (Google Drive archive).
stubmode: output is a copy flow for pipeline testing.commandmode: real inference command from UI (Runtime Settings) orW2L_INFER_CMD.- Variables available in command template:
{video},{audio},{output}{target_json},{track_json},{timeline_json}{project},{app}
Recommended command template:
{app}\\python\\python.exe {app}\\workers\\wav2lip_runner.py --video "{video}" --audio "{audio}" --output "{output}" --target-json "{target_json}" --track-json "{track_json}" --timeline-json "{timeline_json}"
POST /api/video/uploadGET /api/video/meta?video_id=...POST /api/video/probePOST /api/face/selectGET /api/face/selection?video_id=...POST /api/tracking/buildGET /api/tracking?video_id=...POST /api/timeline/saveGET /api/timeline?video_id=...POST /api/job/submitGET /api/jobsPOST /api/output/openPOST /api/job/requeuePOST /api/job/cancelPOST /api/job/deleteGET /api/settingsPOST /api/settings
- LipFlow wrapper code (UI/backend/launcher/scripts in this repository, excluding third-party subtrees): see
LICENSE. Wav2Lip/and related models/checkpoints are third-party and follow upstream terms (including non-commercial restrictions from Wav2Lip project/datasets).- Bundled third-party binaries/libraries keep their original licenses.
Use responsibly and comply with all model/dataset/license and local-law requirements.

