18.8K ·GitHub
What is this?
VideoLingo is an all-in-one AI tool for video translation, localization, and dubbing. It generates Netflix-quality single-line subtitles and can add high-quality dubbing, fully automated in one click.
VideoLingo is an all-in-one AI tool for video translation, localization, and dubbing. It generates Netflix-quality single-line subtitles and can add high-quality dubbing, fully automated in one click.Is it a hosted service, or something I run on my own machine?Something you run locally: it launches a Streamlit interface in your browser and drives the whole pipeline from there, with a Docker option as an alternative.
Something you run locally: it launches a Streamlit interface in your browser and drives the whole pipeline from there, with a Docker option as an alternative.How do I get it running?Install FFmpeg with your system's package manager, clone the repo, and run the one-command uv setup — it installs Python 3.10 and all dependencies for you. On Windows you can also double-click OneKeyStart.bat.
Install FFmpeg with your system's package manager, clone the repo, and run the one-command uv setup — it installs Python 3.10 and all dependencies for you. On Windows you can also double-click OneKeyStart.bat.Do I need API keys for the AI parts, or can it run for free?Normally you configure an OpenAI-compatible LLM (default DeepSeek V4 Flash via OpenRouter), WhisperX, and a TTS service — 302.ai covers all three with a single key. For a free setup, run Ollama locally and use Edge-TTS with no API key at all.
Normally you configure an OpenAI-compatible LLM (default DeepSeek V4 Flash via OpenRouter), WhisperX, and a TTS service — 302.ai covers all three with a single key. For a free setup, run Ollama locally and use Edge-TTS with no API key at all.Where does it still fall short?Background noise degrades transcription, subtitles ending in numbers can get truncated, and weaker LLMs may fail the strict JSON steps — then you delete the output folder and retry with another model. Dubbing is not 100% perfect, multilingual audio keeps only the main language, and per-character voices are not supported.
Background noise degrades transcription, subtitles ending in numbers can get truncated, and weaker LLMs may fail the strict JSON steps — then you delete the output folder and retry with another model. Dubbing is not 100% perfect, multilingual audio keeps only the main language, and per-character voices are not supported.Which systems does it run on?Windows, macOS, and Linux are all supported with per-OS install steps, plus a Docker route that requires an NVIDIA setup. The interface itself is a local Streamlit web app.
Windows, macOS, and Linux are all supported with per-OS install steps, plus a Docker route that requires an NVIDIA setup. The interface itself is a local Streamlit web app.How is it licensed — anything restrictive?Apache 2.0, a permissive license — you can use and modify it freely, including commercially.
Apache 2.0, a permissive license — you can use and modify it freely, including commercially.Do I need an NVIDIA GPU for this?Local WhisperX transcription is where a GPU helps, and the Docker route needs CUDA 12.4 with an NVIDIA driver above 550; Windows NVIDIA users should install CUDA Toolkit 12.6 and CUDNN 9.3.0 first. You can offload transcription to the 302.ai API instead of running it locally.
Local WhisperX transcription is where a GPU helps, and the Docker route needs CUDA 12.4 with an NVIDIA driver above 550; Windows NVIDIA users should install CUDA Toolkit 12.6 and CUDNN 9.3.0 first. You can offload transcription to the 302.ai API instead of running it locally.What actually happens to a video once I hand it over?It downloads the video via yt-dlp, transcribes word-level subtitles with WhisperX, splits them into natural segments with NLP and AI, translates using a custom or AI-generated terminology bank plus a 3-step Translate-Reflect-Adaptation pass, and exports Netflix-standard single-line subtitles — with optional dubbing on top.
It downloads the video via yt-dlp, transcribes word-level subtitles with WhisperX, splits them into natural segments with NLP and AI, translates using a custom or AI-generated terminology bank plus a 3-step Translate-Reflect-Adaptation pass, and exports Netflix-standard single-line subtitles — with optional dubbing on top.Which languages can it work with?Input: English, Russian, French, German, Italian, Spanish, Japanese, and Chinese, with more planned. Translation supports all languages, while dubbing depends on the TTS method you pick.
Input: English, Russian, French, German, Italian, Spanish, Japanese, and Chinese, with more planned. Translation supports all languages, while dubbing depends on the TTS method you pick.How does the dubbing work — can it clone my own voice?Dubbing runs through GPT-SoVITS, Azure, OpenAI, fish-tts, siliconflow, or edge-tts, and customtts.py lets you plug in your own TTS. The demos show voice cloning with Cosy2 and GPT-SoVITS using the author's voice.
Dubbing runs through GPT-SoVITS, Azure, OpenAI, fish-tts, siliconflow, or edge-tts, and customtts.py lets you plug in your own TTS. The demos show voice cloning with Cosy2 and GPT-SoVITS using the author's voice.Why pick it over other subtitle tools?The author positions it against similar projects on three points: Netflix-standard single-line subtitles only (no multi-line), superior translation quality, and a more seamless dubbing experience.
The author positions it against similar projects on three points: Netflix-standard single-line subtitles only (no multi-line), superior translation quality, and a more seamless dubbing experience.What are similar projects?pyvideotrans, voice-pro, Index-Translate, OpenCreator, VoiceStudio, autoclip, seamless_communication, VidBee
pyvideotrans, voice-pro, Index-Translate, OpenCreator, VoiceStudio, autoclip, seamless_communication, VidBeeWhich topics is it in?dubbing, yt-dlp, subtitles, voice-cloning, whisper, localization, speech-to-text, tts, text-to-speech, video-processing, translation, ffmpeg
dubbing, yt-dlp, subtitles, voice-cloning, whisper, localization, speech-to-text, tts, text-to-speech, video-processing, translation, ffmpeg