AI dubbing. Every language, mouths that match.
A dubbing house books a studio, casts voice actors per language and records line by line, and the mouths still say the original words. Our AI transcribes, translates and voices the dialogue, then re-renders the picture so the lips move to the new language. A native-speaking linguist reviews every line, so a series goes out in a new territory in days instead of months.
Picture · SRT
What a dubbing house records line by line, the model does in one pass
We are not a dubbing house with faster engineers. We are an AI lab: the per-line, per-language studio work that fills a vendor’s schedule is done by the model, and people are spent only where judgment matters: the translation and the performance. The result is the same set of deliverables, in days instead of months.
| The job | Traditional dubbing house | Pixazo AI |
|---|---|---|
| Voices | Cast and record voice actors for every language | The original actor’s voice in each language, with consent, or a chosen cast voice |
| Lip-sync | Script rewritten to fit the mouths; picture untouched | Picture re-rendered so mouths match the new dialogue |
| Turnaround | Weeks per language, months per season | Days per episode, per language |
| Cost driver | Studio hours and actors per language | Runtime and languages; one free scene gives the number |
Where people still matter: a native-speaking linguist reviews every translation for meaning, idiom and on-screen fit, and flags performances that need a human voice. Voice cloning only happens with the performer’s written consent.
From one master to every territory
The source stays untouched. Each language comes back as its own audio and picture deliverables, cut to the same timecode as the master.
What actually comes back
Per language, to your distributor’s spec. Tell us the platform and we match its audio layout, loudness and file naming.
| Deliverable | Format | Notes |
|---|---|---|
| Dubbed dialogue stem | WAV, 48 kHz / 24-bit, per language | Clean dialogue only, conformed to the master’s timecode, for your own mix. |
| Mixed language track | Stereo and 5.1, to your loudness spec | Dialogue mixed against your M&E. Tell us the target, for example EBU R128 or the platform’s spec. |
| Lip-synced picture | ProRes, your resolution and frame rate | The master with mouths re-rendered to the new language, per language. Shots without visible speech are passed through untouched. |
| Subtitles & scripts | SRT, VTT, or your timed-text spec; dubbing script | The reviewed translation as subtitles, plus the adapted script per language. |
| Voice-only option | Stems and mixes, no picture change | For territories where a standard dub is fine, or for audio-only and voice-over content. |
The question that decides the scope: does the picture need to change? A voice-only dub is the faster, cheaper line. Mouth-matched lip-sync re-renders every shot with visible speech, so it is quoted per minute of on-camera dialogue.
AI dubbing cannot do every scene
It is very good at dialogue: interviews, drama, documentary, training and factual. It is not a replacement for every performance, and we would rather say so up front.
- Dialogue-led content. Drama, documentary, factual, interviews, courses and corporate film, where clarity and pace matter most.
- Faces in frame. Front-facing and three-quarter shots with visible mouths are where lip-sync makes the biggest difference.
- Many languages at once. Once the source is prepared, each extra language is a fraction of the first.
- Back catalogues. Library titles that were never dubbed because studio costs did not justify the territory.
- Keeping the star’s voice. With consent, audiences hear the original actor in their own language.
- Songs and musical numbers. Sung performance is adapted by people and usually recorded; lyrics are not run through the pass.
- Overlapping dialogue and crowds. Several people talking over each other need the dialogue separated first, which is scoped as audio work.
- Extreme angles and occluded mouths. Profiles, hands over faces and heavy motion blur are passed through as voice-only rather than lip-synced.
- Voices without consent. We do not clone a performer’s voice without written permission; a cast voice is used instead.
- Union and guild productions. Where agreements govern dubbing or synthetic voice, we work to them and say what applies before quoting.
What helps most: the M&E track, a script or as-broadcast subtitles, and a list of names and terms that must not be translated. With those, the first pass is close to final.
What moves the number
We do not publish a rate card, because a voice-only training video and a lip-synced drama with a cast of twelve are not the same job. These are the drivers; the free scene gives you the actual figure.
| Driver | Cheaper | Dearer |
|---|---|---|
| Picture | Voice-only dub | Mouth-matched lip-sync on every speaking shot |
| Source | Clean dialogue stem and M&E | Mixed audio only, dialogue to be separated |
| Content | Single speaker, scripted | Many speakers, overlapping, improvised |
| Languages | Many languages from one source | One language, one-off |
| Volume | A season or a catalogue, priced per minute | Single episodes |
Faster than a rate card: send one scene and the languages you need. We return it dubbed and lip-synced in one language, with a firm per-minute number for the rest. If you are already being quoted elsewhere, tell us the number you need to land under.
Frequently asked questions
Can the dub keep the original actor's voice?
Yes, with the performer's written consent. Without it, a chosen cast voice is used instead.
How does AI lip-sync work?
Shots with visible speech are re-rendered so mouth movement matches the new dialogue. Shots without visible speech pass through untouched.
Who checks the translation?
A native-speaking linguist reviews every line for meaning, idiom and on-screen fit before anything is voiced.
What do I need to send?
The locked picture, the music-and-effects track and the dialogue stem. A script or existing subtitles help; without them the dialogue is transcribed.
Can you dub songs?
Not through the AI pass. Sung performances are adapted by people and usually recorded.
How long does it take and what does it cost?
Days per episode, per language, priced per minute. Send one scene free and you get it dubbed and lip-synced in one language plus a firm per-minute number.
Localization rarely arrives alone
Titles going to new territories usually need restoration and upscaling and an HDR version, and on-screen text and signage may need cleanup and paint. For a single video, try the self-serve AI voice dubbing and AI lip-sync video tools. All of it is inside the Pixazo AI Lab.
Send us one scene and the languages you need.
One scene, no charge, no commitment. You get it back dubbed and lip-synced in one language, the linguist’s notes, and a firm per-minute number for the rest, and you play it to a native speaker before you decide anything.
Send a test sceneTranslation · Original-voice dubbing · Lip-sync · Linguist-reviewed
