Dubbing has been around since roughly 1930, and for all of that time it has worked on one side of a line. You could change everything a viewer hears. You could not change what they see.
Prime Video just stepped over that line, and it did it on a German teen romance.
Amazon has started using AI to alter the mouth movements of actors on screen so they match the English dub, beginning with Maxton Hall, its most watched international original. Both existing seasons are available with the new treatment now, and the third and final season gets it when it arrives on December 9. Amazon says more titles are coming.
Quick facts
- The dubbed voices are still recorded by human voice actors, not generated
- The AI changes the picture, reshaping the original performer’s lip movements to fit the dubbed dialogue
- Maxton Hall seasons 1 and 2 have it globally in English, with season 3 following on December 9
- Amazon says the technology is applied “under creative oversight” to preserve artistic intent
- Amazon’s launch announcement does not use the words consent, face or model anywhere
- Prime Video says the feature will expand to additional titles without naming them
- Viewers can still switch back to the original audio track and subtitles in the player
What is actually being changed
It is worth being precise here, because a lot of the early coverage collapsed two very different things into “AI dubbing.”
Amazon is not generating synthetic voices. Human dub actors still go into a booth and perform the English dialogue, the same as they always have. What is new is the second step. An AI model then takes the original footage and modifies the on-screen actor’s mouth so that it appears to be speaking English.
Raf Soltanovich, vice president of technology at Prime Video and Amazon MGM Studios, framed it as a fix for a familiar annoyance. Amazon’s announcement opens by asking whether you have ever watched a dubbed movie and noticed the actors’ mouths still moving after the dialogue stopped.
That is a genuine problem and this genuinely solves it. It also means the face on screen is no longer only the actor’s.
Why viewers reacted the way they did
The criticism that landed hardest was not about the technology failing. It was about the technology succeeding at the wrong problem.
The objection, stated well
“The issue isn’t that the mouths don’t sync up to the dub, it’s that dubs in general are devoid of much of that emotion. I’d much rather watch with the original audio and translated subtitles than watch foreign actors be puppeted into saying words and making sounds which weren’t part of their performance.”
A second commenter took a lighter route to the same point: “Y’all are gonna make Kung Fu movies suck. One of the most endearing aspects of old-school Kung Fu movies is the melodramatic dub.”
Both of these are arguments about honesty rather than quality. A bad dub is visibly a dub. You know a translation happened, you can mentally discount the voice, and the actor’s face stays available as evidence of what they were actually doing. Smooth the mouth movements and that signal disappears. The scene now presents itself as a performance that never took place.
The Kung Fu joke is doing more work than it looks like too. A whole aesthetic exists because dubbing was imperfect. Fixing the imperfection retroactively across a back catalog would erase something people genuinely like.
The word Amazon did not use
Amazon’s launch page runs to about 357 words. It does not mention consent. It does not mention the actors whose faces are being modified. It does not mention a model, or training data, or who signed off.
What it does say is that the work happens “under creative oversight to ensure the integrity of their artistic vision is preserved,” with “their” referring to creators rather than performers. That is a meaningful distinction in an industry where the director and the actor are represented by different unions and sign very different contracts.
| Question | What Amazon has said |
|---|---|
| Are voices synthetic? | No. Human dub actors still record the dialogue |
| Is the picture modified? | Yes. Mouth movements are resynchronized to the dub |
| Did on-screen actors consent? | Not addressed in the announcement |
| Is it disclosed in the player? | Not described as a labeled or separate track |
| Which titles are next? | “Additional” titles, none named |
| Can you opt out? | Yes, by selecting the original audio track and subtitles |
The silence on consent is the part that will matter commercially, because it is the part organized labor reacts to. Workers have started winning the right to be consulted before AI arrives in their workflow rather than after, and Blizzard’s union just secured exactly that in a contract with Microsoft. A performer’s face is a considerably easier thing to argue about than a workflow.
The original audio track is still in the player. It is just no longer the default for anyone who leaves the settings alone. Photo via Pexels.
Where this sits in a much larger fight
Amazon is not moving into a calm room. Netflix has spent this year absorbing criticism over its own use of the technology, including confirmation that around 300 of its films involved AI somewhere in production. Studios are arguing in public about synthetic performers. Rights holders are in court over whether the models were trained legally in the first place, and Sony and Warner Chappell are seeking $150,000 per song from Anthropic on precisely that question.
Against that backdrop, lip sync looks like the mildest possible AI feature, which is probably why Amazon chose it to go first. Nobody loses a job. Nothing is generated from scratch. The stated benefit, reaching a global audience with foreign language work, is one almost everyone claims to want.
That is also what makes it effective as a precedent. The principle being established is that the photographed image is now editable in post for reasons of distribution convenience. Once that is normal for mouths, the argument for why it should stop at mouths gets noticeably harder to make.
What to do if you would rather not
Open any Maxton Hall episode, go to the audio and subtitles menu, and pick German audio with English subtitles. You get the original performance, the original faces, and the original delivery, which is what most of the people complaining were asking for anyway.
It is a familiar piece of advice by now. The better viewing experience keeps turning out to be the one you have to go and select, whether that is an untouched audio track or, as Christopher Nolan keeps pointing out about bitrate and physical media, a disc instead of a stream.
Amazon’s bet is that almost nobody will bother, and Amazon is almost certainly right. Defaults win. That is the entire reason this is worth paying attention to now, while it is still only a teen romance and only a set of lips.
Sources and further reading
- Amazon: Prime Video introduces lip sync technology on Maxton Hall
- Deadline: Maxton Hall uses Amazon AI to sync lips with human dubs
- TheWrap: Maxton Hall season 3 to use Amazon AI lip syncing
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
