# Voxa

> Voxa is an open-source, MIT-licensed AI video dubbing tool. It transcribes a
> video with Whisper, translates the text with an LLM, synthesizes speech with
> one of four TTS engines, and keeps the dub locked in sync with the speaker
> (anchored placement — zero drift). It runs locally on your own machine.

## Key facts
- License: MIT (open source)
- Author: Akshin Miranov; Publisher: Servoogle
- Runs locally — no SaaS, no account, no telemetry
- Transcription: Whisper (tiny → large, multiple sizes)
- Translation: Google, Ollama (offline), OpenAI, Anthropic
- Text-to-speech: Edge, OpenAI, Piper (offline), XTTS (voice cloning)
- Self-hostable TTS via any OpenAI-compatible endpoint (--openai-tts-base-url)
- Differentiator: anchored placement pins every clip to the source timeline, so
  the dub never falls behind the speaker (unlike naive tools that concatenate
  clips and drift)
- Local operator UI: run `voxa serve` for a browser console (upload, configure,
  watch progress, download)

## Install
- pip install "voxa"
- voxa talk.mp4 --target_lang ru

## Languages
Whisper covers ~100 source languages; target languages depend on the chosen TTS
engine. UI and docs are available in English, Azerbaijani, and Turkish.

## Links
- Repository: https://github.com/akshinmrv/Voxa
- Website: https://voxa.servoogle.com
- Documentation: https://github.com/akshinmrv/Voxa#readme
