A dubbed video should have new words and the same soundtrack. If the whole audio track is replaced, the music and effects go with the old voice. OpenDub replaces only the speech. This is how it keeps the rest.
Separate first
The first step of a dub is separation. Demucs splits the voice from the music and effects. That does two things: the soundtrack survives the dub, and every later step hears a clean voice, which matters for transcribing the words and for cloning the speaker.
Separation runs on your own device, like speech detection, word timing, mixing and rendering. The video file never leaves it.
Two modes: replace the voice, or voice-over
- Replace the voice keeps the music and effects and swaps only the speech.
- Voice-over leaves the original faintly underneath.
Replace the voice is the one to use when the video has background music you want to keep and you do not want to hear the original speaker under the new one.
Mixing the new voice back in
At the end, the new voice goes over the music, matched to the original's loudness. Subtitles are timed to the dubbed voice and burned in.
Each line is also made to fit the moment it replaces. Translations get a character budget sized to how long the speaker took to say the line, and a spoken line more than 15% longer or shorter than the original is tried again, then stretched between 0.75× and 1.3×. So the new speech starts and ends where the old speech did, over the same music.
In the browser
The card at the top of opendub.app dubs videos up to five minutes in your tab. It has a checkbox, Remove the original voice on this device. Ticking it keeps the music and swaps only the speech, and downloads a 172 MB separator once. If the browser has no WebGPU, removing the voice runs on the CPU and can take a long time.
In the app on your computer
The app on your computer has no length limit, and the voice is removed faster. It is also where the free OmniVoice voice runs; OmniVoice's model weights are licensed for non-commercial use only. With Higgs Audio or ElevenLabs you use your own API key and pay that provider directly.
What you get
The app gives you the dubbed MP4, the dub audio as WAV, subtitles in both languages as SRT, and the clip the voice was cloned from. OpenDub is open source under AGPL-3.0, so the separation and mixing steps are code you can read in the repository.
For the whole pipeline step by step, see How OpenDub dubs a video in the speaker's voice.