Why the app, not the browser
- In the browser, on opendub.app or in the browser extension, a dub is limited to five minutes per video.
- In the OpenDub app on your computer, there is no length limit. That is where AI dubbing for long videos runs. There are downloads for Mac (Apple Silicon) and Windows, and a one-line install for macOS and Linux.
How the original voice is kept
The voice is cloned from the densest 6–12 seconds of whole sentences in the original, and every line is then spoken in it. Only clone voices you have the rights to use.
With OmniVoice, the free voice, all of this runs on your computer. With your own Higgs Audio key, the app clones the voice through Boson AI instead, and you pay Boson AI directly.
How the timing holds over a long video
AI dubbing for long videos has to stay in step for an hour, not a minute. OpenDub translates whole sentences, one line out for each line in, and gives each line a character budget sized to how long the speaker took to say it. OmniVoice can speak a line at an exact length, so each dubbed line starts and ends where the original did. Speaking speed is never a setting; it is chosen per line.
Keeping the background music
Choose Replace the voice: Demucs separates the speech from the music and effects, only the speech is replaced, and the new voice is mixed back over the music at the original's loudness. The app separates the voice faster than a browser tab. Voice-over, the other mode, leaves the original faintly underneath. More in Keep background music when dubbing a video.
How long AI dubbing for long videos takes
On a laptop CPU OmniVoice speaks a line in about 15–35 seconds. An hour of speech has hundreds of lines, so expect several hours; the app keeps working while you do something else. The first dub also downloads the voice model once: about 3.6 GB on a Mac in all, 3.3 GB of it the voice.
Fixing a long dub line by line
In a long dub you will want to change a few lines. In the app you can edit a translation, change a line's tone or drag its block on the timeline, then apply. Only the lines you touched are spoken again, not the whole video.
How to dub a long video
- Install the app from opendub.app.
- Choose the video. An MP4, MOV or WebM file on your computer, of any length.
- Choose the language to dub into. One of the 17 the app dubs into; the original language is detected automatically.
- Choose OmniVoice, or Higgs Audio with your own key.
- Choose Replace the voice to keep the music and effects.
- Start the dub, then review it and re-dub any line you want to change.
You get the dubbed MP4, the dub audio as WAV, subtitles in both languages as SRT, and the clip the voice was cloned from. There is no watermark.
Limits of AI dubbing for long videos
- It is slow on a laptop CPU, as above.
- OmniVoice's model weights are for non-commercial use only.
- It does not change the lips in the picture.
- It works on a video file on your computer, not on a link to a video online.
- The Mac download is for Apple Silicon.
For how the app runs everything on your machine, see Local AI dubbing. OpenDub is open source under AGPL-3.0-or-later; the code is on GitLab.
Questions
Can I dub a 60-minute video and keep the original voice?
Yes, in the OpenDub app on your computer, which has no length limit. OmniVoice speaks the translation in a clone of the speaker's voice, for free and for non-commercial use; with your own Higgs Audio key the app clones the voice through Boson AI instead.
What is the length limit?
None in the app on your computer. In the browser, on opendub.app or in the extension, it is five minutes per video.
How long does it take?
On a laptop CPU, OmniVoice speaks a line in about 15–35 seconds, so an hour of speech takes several hours.
Does it keep the background music?
Yes. Replace the voice keeps the music and effects and swaps only the speech. Voice-over leaves the original faintly underneath.
Can I fix a few lines without dubbing everything again?
Yes. In the app, change a line's translation, tone or timing, then apply; only the lines you touched are spoken again.
Is the video uploaded?
No. With OmniVoice the whole dub runs on your computer. With a Higgs Audio key, only speech clips and text go to Boson AI; the video file is never sent.
Is there a watermark on the dubbed video?
No. OpenDub adds no watermark and no logo to the dubbed video. The picture is left as it is; the only thing drawn on it is the subtitles in the new language.