VideoLingo
Streamlit app that transcribes, segments, translates and optionally dubs videos into subtitle files and dubbed video.
Word-level recognition, LLM translation with a terminology step, and optional dubbing in a Streamlit UI.
VideoLingo turns a video or YouTube link into single-line bilingual subtitles and, if you want, a dubbed track. It suits creators who have an LLM API key and want subtitle timing and terminology control more than a plain machine translation.
What it is
It downloads video with yt-dlp, runs word-level recognition and alignment with Qwen3-ASR and Qwen3-ForcedAligner, then segments subtitles with NLP and AI. Translation goes through an OpenAI-compatible LLM with custom and AI-generated terminology, optional reflection and rewriting. Dubbing connects to GPT-SoVITS, OpenAI, Fish Audio, Edge TTS, F5-TTS and others.
Who it's for
It suits subtitlers and educators translating talks, tutorials and interviews between a handful of major languages. It does not suit multi-speaker dubbing, since the README says the dubbing workflow does not assign a separate voice to each speaker.
Key features
Word-level recognition
Uses Qwen3-ASR with Qwen3-ForcedAligner for word timings.
Terminology step
Combines your own glossary with AI-generated terms for consistent translation.
Reflective translation
Direct translation with optional reflection and natural rewriting passes.
Length-limited subtitles
Segments subtitles with configurable length limits for single-line display.
Several dubbing backends
Offers GPT-SoVITS, OpenAI, Fish Audio, SiliconFlow, Edge TTS and F5-TTS.
Resume and HTTP API
Logs progress so tasks can resume, and exposes a local HTTP API for scripts.
Up and running in 3 steps
Our honest verdict
What's great
- Windows one-click installer fetches uv, Python 3.12, dependencies and FFmpeg for you
- Terminology handling keeps domain terms consistent across a long video
- Pause, resume and stop controls with progress resumption help on long jobs
- Works with any OpenAI-compatible LLM endpoint, so you choose the model and price
Where it falls short
- You need an LLM API key and pay that provider; the model must return the structured JSON the workflow requires, and weaker models fail
- Local recognition runs on your hardware; the Docker route needs an NVIDIA GPU, the NVIDIA Container Toolkit and CUDA 12.8.1 or 12.6
- Input languages are limited to eight (English, Russian, French, German, Italian, Spanish, Japanese, Chinese) and mixed-language speech is not guaranteed
- Dubbing timing depends on translation and speech rate, and the README says speed adjustment does not guarantee natural delivery
Pricing & license
Free to self-host under Apache-2.0, plus the cost of your LLM and any paid TTS provider. The hosted service at https://videolingo.io/pricing lists Free at $0/mo (15 minutes, English and Chinese only, 480p, watermark), Starter at $5/mo (60 minutes) and Pro at $30/mo (360 minutes, Fish Audio voice cloning, API access).
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is VideoLingo free?
The code is Apache-2.0 and free to run yourself, but you pay for the LLM API you connect. The project also runs a hosted service at videolingo.io with a free tier and paid plans.
Can it dub with different voices per speaker?
Not automatically. The README lists this as a current limitation of the dubbing workflow.
Which languages can it transcribe?
The README lists eight input languages: English, Russian, French, German, Italian, Spanish, Japanese and Chinese. Dubbing languages depend on the TTS engine you pick.
