AI dubbing for your clips in 12 languages: the voice changes, the music stays
Blog · · Vyexa team · 4 min read
To post the same clip for an audience that speaks another language, you do not need a voice actor. In Vyexa pick a language in the «Dubbing into your language» block on Clip Setup: the AI translates the speech and dubs it in that language. The original speech is turned down under the dub, while music, laughter and the sound of the room stay. Subtitles and the title switch to that language too.
Below: how it sounds, which languages are there, how voices are picked and what it does not do.
12 languages
English, Spanish, German, French, Polish, Ukrainian, Russian, Turkish, Arabic, Japanese, Korean and Indonesian. If the video is already in the language you picked, nothing changes and nothing is counted. When you make more clips from the same video, the dubbing language you used last time is already selected.
Why the background stays
Sound in a video is a mix: voices plus everything else — music, laughter, street noise, game sounds. Simply turning the whole track down under the new voice makes the music dip on every line and come back in the pauses. That «pumping» is what makes cheap dubs tiring to watch.
So we split the sound into two tracks first: voices and background. The original voice goes quiet under the dub, the background keeps playing at full volume the whole time. Where a person speaks but there is no dubbed line, the original stays faintly audible, so you never see someone talking in silence. This is the same idea the big platforms use for their auto-dubbing.
A voice for each person in the video
By default dubbing works in Auto mode for every language: each person in the clip gets a similar AI voice — a man gets a male voice, a woman a female one. Two men talking get two different voices, and the same person keeps the same voice in every clip of the project. Who speaks and whether it is a man, a woman or a child is decided by an AI model that listens to the whole voice, not only its pitch, so a loud or high male voice is still read by a male voice.
Want the same voices every time? On the Dubbing page of your account (Profile → Dubbing) add a language and pick up to three male and three female voices, each with a Listen button. The first man in a clip gets the first male voice, the second man the second one. Leave the male or female list empty and that part is still picked automatically.
What it does not do
- It is a voice-over dub, not lip-sync: lips do not match the new words.
- Music videos are not dubbed — there is singing, not speech. Videos without speech are not dubbed either.
- The translation is done by AI and can miss slang, wordplay or names. Names come out better if you add them to your word list on the My brand page.
- Several people talking over each other, or a loud street, is harder than one clear speaker.
Limits and the API
Dubbed clips per 30 days: Free 3, Creator 30, Pro 100, Business 300. The block shows how many you have left. Through the public API it is the dub_language field, with the same limits — see the API docs.
The voices come from our partner Alvenuri. We have not measured yet how dubbed clips perform compared with the originals, so we give no figure for it — after a dub you can rate it on the results page, and those ratings will tell us.
Try it: paste a link at vyexa.net and pick a language in «Dubbing into your language». Free needs no card.