ElevenLabs has unveiled Dubbing v2, a new AI dubbing model designed to deliver more natural translation and localization of video and audio content, significantly improving emotional fidelity. The model offers preservation of the original speaker's emotion, tone and performance style; translation and adaptation into the most natural expression for the target language; synchronization of generated speech with the timing of the original audio; and automatic voice cloning that replicates the speaker's voice characteristics, pitch and tone. Dubbing v2 analyzes the full performance embedded in the original audio—including emotion, intonation, pacing, pauses and delivery style—and transfers those characteristics into the translated output, recreating a speaker's or character's distinctive voice and emotional expression across more than 90 languages, including English. ElevenLabs believes the technology could substantially reduce the time and cost associated with global content localization by automating much of the traditional workflow. The technology is applicable to entertainment content, educational materials, tourism information, public-service communications, corporate training programs, brand messaging and product introduction videos. Dubbing v2 is currently available through the ElevenLabs user interface platform, and an application programming interface for enterprise customers is expected to be released in the near future.