wtf is violin?

Violin is an open source tool that translates spoken video content into another language while keeping the voice and video in sync. You give it a video file, and it transcribes the speech, translates the text, generates a new voice track that sounds natural in the target language, and merges everything back into a video, with an optional subtitle file. It supports 33 target languages, with especially chosen, natural sounding voices for the 16 most commonly used ones. Under the hood it chains together several pieces: ffmpeg extracts the audio, a speech to text model called Whisper produces word level timestamps, a large language model translates each segment while trying to match a chosen writing style, and a text to speech engine synthesizes the final dubbed audio, before ffmpeg stitches the finished video back together. The project can be used three ways: as a command line tool for translating a single file, as a small web app with its own browser interface and API, or as a skill inside Claude Code that can be invoked directly from a coding session. It also has an in video question and answer feature, where a user can ask about a specific moment in the dubbed video and get an answer based on nearby subtitles and sampled video frames. Configuration is handled through a YAML file, and the underlying AI providers, such as Together AI, OpenAI, Cartesia, and ElevenLabs, can be swapped out. Using it requires an API key from at least one of these providers. Violin is released under the MIT License, a permissive license that allows free use, including commercial use, as long as the copyright notice is kept.