violin

Why we included this project

Violin takes a video in one language and returns a fully dubbed version: it transcribes the speech, translates it segment by segment, synthesizes a native-sounding voice-over, and remuxes the audio back into the file, with optional SRT subtitles. The same pipeline runs as a CLI, a FastAPI web app with a browser UI, or a Claude Code skill, so you can use it from a terminal, host it as a service, or call it from an agent session. Providers are swappable through a single YAML file, letting you mix Together, OpenAI, and ElevenLabs for transcription, translation, and TTS independently. Beyond straight dubbing, it includes a natural-language voice picker, six tone profiles for different audiences, and in-video Q&A that answers questions about specific moments using nearby subtitles and sampled frames. Teams localizing lectures, tutorials, or marketing clips will want to check the alignment and voice quality against their own workflows.