VideoLingo 🧩 Workflow Open source

What it is

VideoLingo by Huanshere (Apache-2.0) produces high-quality translated subtitles and optional dubbing. It uses WhisperX for word-level timestamps, splits subtitles with NLP and an LLM so lines read naturally, translates in a translate-reflect-adapt loop with terminology control, and can dub with several TTS engines. It runs as a Streamlit web app locally or in Docker. The focus is on subtitles that read like professional work: short single lines, timing that follows speech, and consistent terms across a whole video, which generic auto-caption tools often miss.

Key features

• WhisperX word-level transcription and alignment
• Netflix-standard single-line subtitle segmentation
• Three-step LLM translation with custom terminology
• Dubbing with Azure, OpenAI, Fish TTS, CosyVoice2, GPT-SoVITS, Edge TTS or F5-TTS
• One-click YouTube download and processing
• Streamlit UI and Docker image

How to install

Clone with git clone https://github.com/Huanshere/VideoLingo.git, then run uv run --no-project --python 3.13 setup_env.py to set up the environment (FFmpeg 7 shared libraries required), or build the Docker image with docker build -t videolingo ..

Pricing and rights

Free under Apache-2.0. You pay for the LLM API used for translation and any cloud TTS; WhisperX and several TTS engines can run locally. Dubbing someone else's voice needs their consent.

Who it is for

Creators and educators who want professional-looking translated subtitles for their own videos, with dubbing as an option.

Verdict

Among the best open-source options for subtitle quality thanks to careful segmentation and translation. It needs a GPU for fast transcription, and the setup has specific FFmpeg requirements.