Descript Adds Generative Lip-Sync to Its AI Video Translation and Dubbing Feature

Descript's video translation and dubbing tool now goes beyond traditional rotoscoping by using generative AI to reconstruct the lower half of a speaker's face to match the mouth movements, pacing, and sounds of a newly dubbed language. The system encodes the original video into a latent space, regenerates the mouth region using the translated audio while preserving lighting and speaker identity, and then blends the result seamlessly with the untouched portions of the frame.

A hands-on review tested the feature across multiple languages — two varieties of French, German, and Greek — and found the lip-sync quality notably accurate. The reviewer, a certified translator, notes that while the AI-only translation is usable for quality assessment, curated human-native-speaker review (available on Descript's Business and Enterprise plans) remains preferable for production-ready multilingual content.