AI dubbing: how to translate a video into another language

AI dubbing translates what is said in a video and voices it in another language, timed to the picture. How it works, when subtitles do, and consent.

· Founder of Sharpy and Threadify18 October 2026 · 6 min read
A young woman speaks into a microphone in a small dubbing studio while a man at the mixing desk behind her listens on headphones.

The short answer

AI dubbing transcribes what is said in a video, translates it, voices the translation in another language, times each line to the picture and mixes it under the original music and effects, sometimes re-syncing the mouth too. Subtitles are the lighter alternative. Sharpy does both on your Mac: it transcribes in the language spoken, translates into a subtitle track per language, and dubs in a voice fitted to each line's length and pauses, with the mouth re-synced and the sync measured per language.

What is AI dubbing?

AI dubbing is replacing a video's speech with a translated, synthetic voice, made by software instead of a cast in a studio. It runs through the same stages a human dub does, each one a place where it can go wrong:

Stage What happens What to check
Transcribe Every word said is written down, with its timing Names and terms heard correctly
Translate Each line is put into the new language Meaning, not just words, and length
Voice A voice speaks each translated line Pronunciation, and whose voice it is
Time Each line is fitted to the time the original took No line spills over the next shot
Mix The new voice goes over the original music and effects The original voice is gone, the rest stays
Lip sync The mouth is changed to match the new words Close-ups, where a mismatch shows most

The hardest stage is timing. Translations run long (a sentence in Spanish or German is often longer than in English), so a good dub rewrites a line to fit, rather than gabbling it.

How do you dub a video into another language?

  1. Start from a clean transcript. Fix every misheard name and term first; a mistake here is copied into every language.
  2. Translate line by line, keeping each line's meaning and aiming for its length.
  3. Voice each line in the target language, with the speaker's permission if you use their voice.
  4. Fit each line to the picture. Shorten the wording before you speed up the voice.
  5. Mix it over the music and effects, without the original voice. This needs the dialogue on its own track or stem.
  6. Check it with a native speaker, then publish it as its own audio track, labelled as dubbed.

On YouTube, creators with access to advanced features can add their own dub as an extra audio track: an audio-only file roughly the same length as the video, uploaded in YouTube Studio's Languages section. YouTube also dubs eligible videos itself, marking them "auto-dubbed" in the description and letting creators review a dub before it is published.

Should you dub or translate with subtitles?

Subtitles first, in most cases. They are cheaper, quicker to check, and keep the speaker's own voice. Dub when your audience would rather listen than read:

Subtitles Dubbing
Effort Translate and time the text Translate, voice, time and mix
The speaker's voice Kept Replaced, unless you clone it with permission
Best for Talking heads, interviews, anything with personality Tutorials, children's video, anything watched while doing something else
Watching Viewers read Viewers listen
Risk A clumsy line is read once A clumsy line is heard in a stranger's voice
An older couple on a low sofa watch a person talking on a television in their living room in the evening, cups of tea on the table in front of them.
A dub lets people listen in their own language; subtitles ask them to read but keep the speaker's own voice.

How does Sharpy translate a video?

On your Mac, in four steps. It transcribes in the language actually spoken, which you can name or let it hear; if it is less than half sure which language it is, it asks. Two speech engines vote on every word in 25 European languages, and measured on Google's FLEURS recordings on the base Mac mini M4 with 16 GB, it gets 3.81% of words wrong in Spanish, 5.35% in French and 3.56% in German. Transcribing video to text on a Mac has the detail.

It then translates each caption whole, re-breaks it into readable lines, keeps its timing, and writes each language as its own subtitle track in the video, plus .srt and .vtt files beside it. If your Mac does not have a language yet, the window downloads it when you ask. Every translated track is labelled as machine translation, for a reason: "Pin your plugins" once came back in Spanish as "catch your plugins" and in French as "paste your plugins". Sharpy also lists every caption that breaks its reading rules once translated, because translations run longer. What an SRT file is explains the subtitle files.

How does Sharpy dub a video?

It dubs each line in a voice fitted to that line's length and pauses, so no line spills over its shot, and it re-syncs the mouth to the new words, measuring the sync in each language rather than assuming it.

Two things make the mix work. Sharpy knows who is speaking when, so each speaker's lines are handled as theirs. And it writes undipped stems: each ducked element, such as the music bed, at a steady level with no dips, for re-setting under a voice in another language, beside stems that add up exactly to the original mix. Its line-fitting tool, which fits a new reading of a line to the original's timing with each word starting where the original's did, is the same one it uses for an AI voice-over.

Your own voice can carry the dub too, with your recorded permission: overdub says words in the speaker's own voice and matches the mouth, and the export carries Content Credentials and an invisible watermark marking it as altered. Sharpy's own synthetic voice is always declared in the description of a piece that carries it.

Yes, for your own videos, with disclosure where the rules ask for it. The questions are whose video it is and whose voice it is.

  • Someone else's video. A translation is a derivative work, and the US Copyright Office's guidance is that only the copyright owner may make or authorise one. Dub only what you own or have permission to translate.
  • Someone else's voice. Many US states protect a person's voice as part of their likeness, and the US Copyright Office's July 2024 report on digital replicas recommended a federal law against unauthorised ones. Clone a voice only with its owner's consent.
  • Disclosure. The EU AI Act's transparency rules (Article 50) apply from 2 August 2026: synthetic audio has to be marked as such in a machine-readable way, and a deep fake has to be disclosed. YouTube lists cloning your own voice for dubs among the uses that need no disclosure, but asks you to disclose a real person seeming to say something they did not.

Frequently asked questions

01

What is AI dubbing?

AI dubbing is software that translates a video's speech and speaks it in another language, timed to the picture and mixed over the original music and effects. Some tools also change the speaker's mouth to match.

02

Is lip sync necessary in a dub?

Not always. It matters most in close-ups of a speaker's face; over B-roll, screen recordings or wide shots, a well-timed voice is enough. Sharpy re-syncs the mouth and measures the result per language.

03

How accurate is AI video translation?

Good enough to start from, never good enough to publish unread. Sharpy labels every translated track as machine translation; have a native speaker read it first.

04

Can I dub someone else's video?

Only with the copyright owner's permission, because a translation is a derivative work. Use their voice only with their consent as well.

05

Does YouTube dub videos automatically?

For eligible channels, yes. YouTube marks those videos "auto-dubbed" in the description, and creators can review a dub before it goes out or upload their own dub as an extra audio track.

Sources

Checked on 18 October 2026.

hopeengineer

Written by

hopeengineer

Software architect and founder of Sharpy and Threadify. Builds scalable systems and AI tools, and writes here about how Sharpy understands and edits video.

Sharpy opens on 1 November 2026.

The world's first context-aware agentic video editor, on your Mac. One plan, unlimited edits.

Get early accessSee the live price