Tools · Model Watch
Dubbing v2 promises to keep the original delivery. Check how you can fix a line.
ElevenLabs’ model supports more than 90 languages, but its in-app output does not have an edit button.
Reporting updated Sep 23, 2026
THE TAKEAWAY
Try one difficult scene before dubbing a whole project, and find out how you would correct a line that misses.
What Happened
ElevenLabs launched Dubbing v2 on May 28. It says the model uses the original speaker’s performance to guide translated speech, carrying over tone, pacing and emphasis across more than 90 languages.
The current app workflow lets you upload audio or video, choose target languages and adjust speaker similarity before generating the dub. The company also offers a separate professional service with human translation, casting and mixing.
The Correction Route
ElevenLabs’ documentation says v2 dubs are automatic, with no in-app option to edit their content. Transcript editing and audio regeneration are available through the API on Enterprise plans.
The Speaker similarity setting has its own tradeoff: higher values favor resemblance to the original speaker, while lower values allow more natural delivery in the target language, according to the guide.
Why It Matters
If a translated line sounds wrong, the next job is fixing it. Knowing that route before starting a long film helps a small team plan the review and corrections, rather than discover the limit after generating every language.
Start with a Difficult Scene
Choose a passage with a joke, a pause or a change in emotion. Ask a fluent listener whether the meaning and performance survive. Then try correcting one line; a promising first listen should not be the only test.
.webp)