From the archive
ElevenLabs launches Dubbing v2 for translated video audio
Dubbing v2 uses the original performance to guide translated speech, with API access planned after the initial release.
ElevenLabs introduced Dubbing v2 on May 28, 2026. The model used the original speaker's performance to guide translated speech, including its timing, tone, and delivery.
Translation with the performance attached
The system conditioned generation on the original audio rather than only a transcript. Its synchronization-aware translation adjusted phrasing and pacing for the target language. The release covered more than 90 languages and was aimed at videos, advertising, and other localized media.
Product access first
Dubbing v2 launched in ElevenCreative and through ElevenProductions. The latter combined the model with professional localization services such as translation, voice casting, and mixing. API access was still planned separately in the launch announcement.
The release offered a route for localizing existing recordings while keeping the original performance involved in the generated result.