In a major breakthrough for AI-driven video localization, industry pioneer ElevenLabs has officially launched Dubbing v2.

While first-generation AI dubbing tools focused strictly on translating words and slapping a generic synthetic voice over a video, ElevenLabs’ new foundational model introduces a massive architectural shift: it conditions the translation directly on the source performance rather than a flat text transcript. This allows the original speaker’s exact tone, emotional nuance, delivery, and timing to carry over flawlessly into every target language.

The Technical Leap: Translating the Performance

Dubbing v2 automates what used to require an entire pipeline of audio engineers, translators, and voice actors. The tool handles transcription, translation, automatic voice cloning, and audio-to-video synchronization in a single loop.

The primary upgrades anchoring the v2 model include:

• True Emotional Continuity: For the first time, if a speaker laughs, gasps, or sounds angry in the original video, that specific vocal inflection is natively preserved in the dubbed output.
• Zero-Setup Voice Cloning: The model automatically samples the speaker’s original voice, analyzing identity, pitch, and timbre. It then regenerates that exact voice clone speaking the new target language fluently without requiring manual training data.
• 90+ Languages and Accents: The system’s linguistic footprint has been dramatically expanded, allowing creators and enterprises to instantly map content naturally to localized native audiences globally.
• Meaning-Based Phrasing: Instead of rigid word-for-word translation, Dubbing v2 dynamically adapts idioms and phrasing so the script sounds natural to native speakers while maintaining the original video’s timing constraints.

The Legal Fine Print: Restricted Commercial Uses

While ElevenLabs is heavily pitching the model to creators, corporate marketers, and localization studios, the company has implemented strict legal boundaries within its newly published Dubbing v2 Model-Specific Terms.

Unless an enterprise client executes a specialized, separate written agreement or order form directly with ElevenLabs, the output from Dubbing v2 is strictly prohibited from being used in:

• Feature films
• Television programs or broadcast series
• Scripted streaming productions and video-on-demand (VOD) platforms
• Theatrical movie releases

This restriction applies regardless of the final distribution channel (cable, streaming, or broadcast), effectively drawing a legal line between web-based content creation, localized corporate marketing, and traditional high-end Hollywood entertainment.

Positioning within the ElevenLabs Ecosystem

The launch of Dubbing v2 rounds out an incredibly busy infrastructure cycle for ElevenLabs, which has rapidly diversified its product lines beyond its core web-based text-to-speech engine.

The Dubbing v2 launch follows the recent releases of Music v2 (their studio-grade music generator) and Scribe v2 (their highly accurate speech-to-text transcription engine), firmly positioning the firm as a full-stack audio infrastructure layer for developers and modern content workflows.