How Descript engineers multilingual video dubbing at scale

Descript, a video editing platform, has leveraged OpenAI's reasoning models to automate multilingual dubbing at scale, according to the OpenAI Blog. The challenge was to localize large video libraries without breaking precise timing or altering meaning—a common pain point in AI-driven translation. By integrating reasoning models, Descript’s system can analyze context, preserve speaker intent, and adjust translations to match lip movements, bypassing the need for manual frame-by-frame adjustment. This approach allows creators and enterprises to efficiently expand their reach across languages while maintaining quality. For developers building AI workflows, this case demonstrates how reasoning models can enforce complex constraints—like timing and semantics—that simpler translation APIs often miss.