ElevenLabs presents Dubbing v2, its new dubbing model

ElevenLabs launches Dubbing v2, an AI model that translates voices in 90+ languages by directly preserving the original tone, emotion, and performance.

Dubbing v2's unique feature stems from its method: rather than starting from a transcript, the model directly conditions itself on the original performance, which allows it to transfer the tone, emotion, and pace of the voice into the target language. The publisher describes this as the main pitfall that AI dubbing had not yet been able to address.

The model also adjusts the phrasing for a natural output in over 90 languages, with a translation logic attentive to synchronization, so that the beginnings and ends of sentences coincide with the original. The firm points out that professional dubbing can cost several hundred dollars per minute, whereas the tool operates in a fully automated manner.

Access is available now in ElevenCreative and via the ElevenProductions offering, which combines human translators, voice casting, and mixing, with the model handling the audio portion. Creators can also apply to the Creator Dubbing Partner Program, which offers reduced-rate access in exchange for publishing dubbed content mentioning ElevenLabs. For seven days, a free trial is also offered, ranging from one to thirty minutes of usage depending on the plan, with API integration planned for later.