Audio AI startup ElevenLabs said on Aug. 10 it will provide its new dubbing AI model, Dubbing v2, through ElevenAPI. The model reproduces original emotional expression and performance.
Dubbing v2 was first unveiled as a UI version on May 28. The API release enables developers and companies to connect and integrate it directly into their products and workflows.
The company said Dubbing v2 is based on an audio-to-audio architecture that converts speech directly into speech. Unlike a pipeline of recognition, translation and voice synthesis using cloned voices, it generates output directly from the input voice, allowing it to better reflect original emotion and performance.
ElevenLabs plans to expand the Dubbing v2 model, targeting corporate customers such as content creators, media companies and enterprises seeking to provide high-quality content for global audiences.
The 92 languages supported by Dubbing v2 include Korean and Japanese. The company said it expects the model can be used in both directions, including high-quality multilingual dubbing for K-content entering global markets and Korean dubbing of leading overseas content.
Chi-gyu Hwang
delight@d-today.co.kr