Stock voices, never your own. One clip per caption, on its own track.
Dub it
Open Text › Captions. Under Languages, open a language's ⋯ menu.
Press Dub this language.
Pick a Voice.
Press Dub. The first time the button reads Download 341.7 MB and dub: the voices download once.
The result says, for example, "Dubbed 3 lines in Spanish", and the language's row says Dubbed.
Press Undo in the result, or Ctrl+Z, to take the whole dub back.
Where the voice comes from
This computer
AI cloud
Cost
Free
Credits
Works offline
Languages
English, Spanish, French, Italian, Brazilian Portuguese, Hindi
23 languages, including German, Dutch, Japanese and Korean
How
Made on your computer
Ask the AI writes the request in the chat; the AI buys the voices within the edit's limit and asks first for a step of 10 credits or more
Marked as an AI voice
Where
What it says
Timeline and Media
AI voice on the dub file, and on each dub line wide enough to show its name
Your exported file
Inside the file: it contains an AI-generated voice (the EU AI Act asks for this)
Some sites remove that information when you upload. When you post a dubbed video, say the voice is AI-made.
What happens to the original sound
Choice
What you hear
Lower it (default)
The original dips under each line
Replace voice
The separated voice goes silent; music and effects stay. Shown after Separate sound
Keep it
The original plays as it is
8 more details
Translate first: a new language is a copy of the original captions. If it is still Not translated yet, the dub panel warns you and offers Translate with the AI; dubbing anyway reads the original words. See Subtitle languages.
A line that is longer than its caption is sped up a little without changing the pitch (up to 1.2×). A line that still doesn't fit is left out, and the result lists it. Shorten that caption and dub again.
Dub again replaces the earlier dub of that language. Remove dub takes it out and brings the original sound back exactly.
The dub lines are ordinary audio clips: move, trim, change the volume or mute them like any other sound.
The voices on this computer are Kokoro (Apache 2.0) with eSpeak NG for pronunciation (GPL 3.0), downloaded from their makers' own releases.
French has a female voice only on this computer. European Portuguese and Japanese use the AI cloud. Mandarin can't be dubbed yet.
With strict privacy for AI-generated media, the AI cloud voice is off. Its lines go through OpenRouter to the voice model's maker: see What is sent to the cloud.
Synthetic voices can mispronounce names, numbers and abbreviations. There is no lip sync.