← Back to blog
AiAbout 6 min read

ElevenLabs Put Lip Sync Inside Dubbing, and a Two-Job Line Item Disappeared

Published Nov 14, 2026
ElevenLabs Put Lip Sync Inside Dubbing, and a Two-Job Line Item Disappeared

On November 12, ElevenLabs added Dub & Sync to ElevenCreative. You upload a video, pick the languages you want, and get a dub back with the lips fitted to the new voice track. One run, no separate step. The announcement carried no prices and no count of supported languages.

That is a small sentence for a change that removes an entire line from a production budget.

How the work was split, and why

Localizing a talking-head video into five languages has been a five-stage job for as long as the job has existed. Translate the script. Cast or synthesize voices. Record or generate the audio. Fit the mouth to the new audio. Conform and check.

The fourth stage is where the money went, because it is where the two halves of the problem disagree. A translation changes the words and the length. German runs longer than English. Japanese runs shorter and lands stress in different places. The voice track arrives with a rhythm that the original performance was never built around, and somebody has to reconcile a mouth that is forming English phonemes with a track that is speaking Portuguese.

The usual answer was to hire a second vendor. Dubbing companies handled voice. Lip sync companies handled the face. Each had its own queue, its own file formats, and its own idea of what "delivered" meant. Between them sat a conform step that nobody billed for and everybody paid for.

Merging stages four and five removes a supplier, which is a larger change than a feature.

What the interface implies

Dub & Sync takes the video as the input and returns the video. That framing matters more than the model.

When lip sync is a separate product, you feed it a video and an audio file, and it returns a video with a new mouth. The unit is a fixed asset, and any change upstream means starting over. When lip sync lives inside dubbing, the unit is the localization job, and the system can hold the script, the voice, and the face in one place while it works.

The second framing allows something the first cannot: fitting the translation to the performance rather than the other way round. If you know the mouth has to carry a phrase in a fixed number of frames, you can rewrite the phrase. Professional dubbing has done this for decades under the name "adaptation", and it is the difference between a dub that sounds translated and a dub that sounds written.

ElevenLabs did not say whether Dub & Sync does adaptation, and that is the question to ask. Without it the product is a faster handoff. With it the product is a different job.

The quality gate that used to exist by accident

There is an underrated benefit to splitting a job between two companies. The handoff is a checkpoint. Somebody looks at the audio before it goes to the face, and somebody looks at the face before it ships. Two vendors means two moments where a human being could notice that the third line is wrong.

Collapsing the pipeline into one run removes both. Every integrated tool makes that trade. The check should move rather than disappear.

The checks that matter in a dub are specific. Does the voice match the original speaker's age, register, and pace? Does the mouth movement land on the right syllables at the start and end of each phrase, which is where sync failures are most visible? Does the emotional arc survive, or does every language come out at the same mild enthusiasm? The last one is the failure mode of synthetic voice in general and it is the one no lip sync metric will catch.

Who this actually changes things for

Three groups, with different math.

Agencies localizing a client's campaign into eight markets currently build a budget with a dubbing line and a lip sync line. Consolidating them cuts both cost and calendar, and the calendar matters more than the cost because campaign dates do not move.

Course and training video producers are the second group. They re-record constantly, and every re-record triggers a re-dub in every language. A pipeline where the video is the input and the output makes the fifth revision cheap instead of nearly as expensive as the first.

The third group is the one that should worry somebody. Anyone making performance-driven content, where the actor's face is the product, now has a tool that can rewrite what that face says in a language the actor does not speak. The output is convincing enough that the question stops being technical.

The consent problem underneath

A dub is now an edit to a person's mouth. That sentence has been true since lip sync tools appeared, but the tools used to be separate, awkward, and easy to describe as post-production. Folded into a dubbing product that takes one upload and returns one file, the same operation becomes routine.

Most standard talent agreements cover dubbing into other languages. Fewer cover synthetic re-animation of the performer's face, and fewer still address what happens when the voice is synthetic too. A contract written in 2023 that granted "the right to dub" probably did not imagine the mouth being regenerated frame by frame to match a cloned voice.

I am not arguing that the tool is the problem. The problem is that the paperwork has not caught up, and the paperwork is what a studio checks before it ships. If you are producing talent-led content for multiple markets, the practical step this month is one paragraph added to the release: consent for synthetic voice and synthetic facial re-animation, per territory, with a term.

Where the price lands

ElevenLabs declined to publish pricing for Dub & Sync, which is normal for a launch week and also the only number a buyer needs.

The way to estimate it is to price the thing it replaces. A minute of professional dubbing into a mid-size language market runs into the hundreds of dollars once you include adaptation, casting, studio time, and QC. A minute of lip sync from a specialist vendor runs lower but adds a second queue. If Dub & Sync lands between the two, it wins on calendar. If it lands near the lip sync price, it wins on everything.

The bet the product makes is that most buyers were never paying for craft. They were paying for coordination, and coordination is now software.

That is the same bet Adobe made by putting partner video models inside Premiere, and the same one Krea made by putting house rules into agent memory. The models are becoming interchangeable. The handoffs are not, and whoever removes a handoff keeps the customer.

What to test first

Take one finished video with a performance you already approved. Run it into three languages you can actually evaluate, ideally with a native speaker in the room. Watch the first two seconds of every phrase, which is where lip sync breaks. Then watch the last two.

If the mouth holds at the boundaries, the tool is usable. If it holds everywhere except the moment a sentence changes direction emotionally, you have found the part that still needs a human, and you know what to budget for next quarter.

Related articles