Why does a translated product spokesperson still have the original-language mouth movements?

Identify the difference between language timing and visible mouth shapes

The official ElevenLabs Dubbing documentation states that lip syncing is not part of Dubbing. Other ElevenLabs Image & Video, Flows, or Studio workflows may involve separate third-party model capabilities, but those should not be treated as an automatic Dubbing feature.

Translated speech may have a different word order, syllable pattern, and duration from the original performance. Even when its overall timing fits the scene, the visible mouth can still form shapes associated with the original language.

For merchants, the relevant question is not whether translation has finished. It is whether the advertisement's approved creative brief requires viewers to see a person visibly speaking the translated words. A product close-up and a front-facing spokesperson have different visual obligations.

Choose between product narration and speaking-face delivery

Branch A — product-only shots or voiceover: if the advertisement shows tabletop demonstrations, packaging, hands, or product details without a visible speaking mouth, translated narration and scene alignment may be sufficient. Do not commission a separate lip-sync process merely because the language changed.

Branch B — approved front-facing spokesperson: if a recognizable person must visibly deliver the translated message, plan a distinct visual-synchronization stage. Confirm authorization to use the person's likeness, the available source footage, target audio, and the current requirements of any proposed model or production method.

Illustrative exercise: a coffee-machine advertisement alternates between a presenter explaining the controls and close-ups of the buttons. Mark the presenter shots as potentially requiring mouth synchronization. Mark the button demonstrations as narration-only candidates. This shot-level distinction avoids treating the entire advertisement as one expensive visual problem.

Build a shot inventory before selecting a visual workflow

For each shot, record whether a mouth is visible, whether the person appears to be speaking, whether their identity must remain recognizable, and whether a product is held or demonstrated. Also record the intended translated line and its approximate placement.

A separate synchronization workflow must be evaluated against actual footage and target-audio duration. Do not assume a particular provider supports every face angle, multiple people, obstruction, or hand-held product interaction. Capability, permissions, and required inputs must be established before committing to that branch.

If the merchant declines visual generation, an alternative is to revise the creative brief toward product B-roll and narration, subject to approval. That is a different advertisement treatment, not evidence that the original speaking-face version has been synchronized.

Hand off two independent approvals

First, the language reviewer should approve translated claims, product terminology, intended meaning, and approximate relationship to the scenes. Second, the visual reviewer should examine only shots where speaking-face synchronization is contractually required.

The visual checklist should cover mouth-to-word correspondence, face identity, facial continuity, shot timing, and the appearance of any product being held. A frozen-frame comparison or a few coincident mouth openings cannot establish continuous synchronization.

Record unresolved problems by shot identifier and assign an owner for correction. Never label a translated advertisement as lip-synced solely because its audio is complete. Final acceptance must reflect the approved creative branch and actual review evidence, not an assumed model accuracy or an unperformed generation test.

Illustrative exercise: spokesperson shot and language handoff

These are illustrative exercise rows. Save a copy and record your own observations; the example is not an executed acceptance test.

Download review worksheet (CSV)