ElevenLabs Dubbing v2 keeps the drama across languages

AI dubbing has always had a fatal flaw: the words get translated, but the soul gets lost. An actor can be screaming in tears on screen, while the dubbed voice sounds like a GPS giving turn-by-turn directions.

ElevenLabs just launched Dubbing v2, a brand-new dubbing model built specifically to fix that disconnect.

Translating the acting, not just the words

According to ElevenLabs, Dubbing v2 carries over the emotion and performance of the original content into every supported target language. The company claims this marks the first time a dubbing system can fully retain subtle emotional nuances from the source take across multiple tongues.

Instead of just matching voice timbre and syncing lips, the model aims to capture the sighs, anger, excitement, and pacing that made the original performance work in the first place.

Why it matters

For filmmakers, game studios, and YouTubers eyeing global audiences, performance is everything. Traditional dubbing requires hiring local voice actors to replicate emotional beats from scratch, which is slow and wildly expensive. If Dubbing v2 delivers on its promise, a solo creator could theoretically localize an emotionally complex video for global distribution in minutes.

We think the AI voice race has officially moved past natural pronunciation and straight into acting capability. Focusing on emotional fidelity is a smart play by ElevenLabs to lock in media companies. The real test, however, will be cultural nuance: what sounds dramatic in one language can easily feel overacted or bizarre in another.