ElevenLabs Launches Dubbing v2: AI Translation Across 90+ Languages That Keeps the Original Voice

Summary: ElevenLabs shipped Dubbing v2 on May 28, 2026 — a model that skips the traditional ASR→translate→TTS pipeline and instead directly conditions on the original audio performance to produce translated speech that sounds like the same speaker.

Key Facts

• Novel approach: Conditions directly on source audio (not a text transcript), capturing emotion, intonation, pacing, and pauses before generating the translated output
• Language coverage: 90+ languages and accents with automatic voice cloning that replicates speaker pitch, tone, and timbre
• Sync: Generated speech automatically aligns to the original video or audio timing
• Availability: GA as of May 28, 2026 via ElevenCreative and ElevenProductions; free tier includes up to 30 minutes on Creator+ plans

Why It Matters

Content localization has long struggled with the tradeoff between translation accuracy and preserving speaker identity. Dubbing v2 collapses both problems into a single model, making it feasible for individual creators, podcasters, and enterprise L&D teams to publish in 90+ languages at near-zero marginal cost — a significant shift for global content distribution economics.

Read More

• Introducing Dubbing v2 — ElevenLabs Official Blog
• ElevenLabs Dubbing v2 Launches With 90+ Languages — aiHola