jianchang512/pyvideotrans
AUTOMATIC VIDEO TRANSLATION WITH DUBBING
Open-source tool for full-cycle video translation: speech recognition, subtitle translation, speech synthesis, and video assembly. Supports local models and cloud APIs for ASR, LLM, and TTS.
// KEY FEATURES
- Full pipeline: ASR → translation → TTS → video assembly
- Speaker diarization and multi-role dubbing support
- Voice cloning: F5-TTS, CosyVoice, GPT-SoVITS
- Local (Ollama, Faster-Whisper) and cloud APIs
// DEEP DIVE: HOW IT WORKS
pyVideoTrans automates the full pipeline: ASR (Faster-Whisper locally or cloud APIs) → subtitle translation via LLM (DeepSeek, GPT, Claude, Ollama) → TTS (Edge-TTS, OpenAI, cloning via F5-TTS/CosyVoice/GPT-SoVITS) → video assembly. Each stage allows pause and manual editing.
Key feature — speaker diarization and multi-role dubbing with distinct voices. Includes utilities: vocal separation, A/V synchronization, subtitle merging. CLI for batch processing and WebUI for remote access. Supports dozens of ASR/LLM/TTS providers, mix of local and cloud models.
Windows: ready .exe (GPU requires CUDA 12.8, cuDNN 9.11, NVIDIA only). Source deploy: Python 3.10, FFmpeg in PATH, uv. Local models need NVIDIA GPU, otherwise slow. Cloud APIs require keys.
// USE CASES
Best for: Content creators and video localizers for multilingual audiences
- Translate educational video into 10 languages preserving speaker voices
- Create role-marked subtitles for podcast and dub with different voices
- Localize marketing video via local models without sending data to cloud
// QUICK START
brew install libsndfile git python@3.10
brew uninstall --ignore-dependencies ffmpeg
brew tap homebrew-ffmpeg/ffmpegCommands are taken from the repository README — check the current version on GitHub before running.
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram