Huanshere /VideoLingo

About VideoLingo

Hand VideoLingo a video file or a YouTube link and it writes the subtitles. Each word gets a timestamp, then it splits the text into lines and translates them. Dubbing is optional, and it builds a voice track too. The catch: most LLM, speech and voice providers need paid API keys.

Best for: Anyone who wants subtitled or dubbed versions of videos they already have.

• Downloads YouTube videos with yt-dlp
• Times every word with Qwen3-ASR and Qwen3-ForcedAligner
• Splits subtitles into lines with length limits you set
• Keeps names and terms consistent with a custom glossary
• Dubs the audio with GPT-SoVITS, OpenAI, Edge TTS and more
• Pauses, resumes or stops a job from the Streamlit window

1. 1
   Windows: download the source zip from the latest release, extract it, then double-click OneKeyStart.bat.
2. 2
   On any OS: the command below then uv run start.py.

   git clone https://github.com/Huanshere/VideoLingo.git && cd VideoLingo

   Copy
3. 3
   Next time, run uv run start.py again from the VideoLingo folder.
4. 4
   Optional Docker on Linux with an NVIDIA GPU: docker build -t videolingo . then the command below.

   docker run -d -p 8501:8501 --gpus all videolingo

   Copy
5. 5
   For scripts, start the local API with uv run start.py --api (endpoint docs at localhost:8000/docs).
6. 6
   Enter your API URL, key and model in the Streamlit sidebar (the browser window the app opens).