ByteDance's Seed Audio 1.0 brings text-to-music generation to its AI creative suite

ByteDance released Seed Audio 1.0 on July 20, 2026, marking its entry into AI music and sound generation. The foundation model supports three core capabilities: text-to-audio, where users describe a track and get a generated composition; audio continuation, which extends existing audio clips; and sound-effect generation for adding foley, ambience, or UI sounds. Seed Audio 1.0 is designed to slot directly into the workflow of Seedance 2.5, the company's video model that produces 30-second 4K clips in a single pass with support for 50 mixed multimodal references and region-based editing. Together, they give creators a unified pipeline for producing both visuals and synchronized audio without juggling separate tools.

Seed Audio 1.0 is not an isolated release. It arrives amid a rapid six-week cascade: Seedance 2.5 previewed at the Volcano Engine FORCE conference on June 23, then went live in Dreamina, Jimeng, and CapCut on July 31. Doubao, ByteDance's chatbot with 345 million monthly active users, provides the distribution layer. The company is also reworking its first-generation AI glasses after scrapping the original design. This is a coordinated push to build a full-stack creative operating system—text, image, audio, video, and hardware—all feeding into the same engine. For creators, Seed Audio 1.0 means soundtracks and effects can be generated directly from prompts, aligned with AI-generated footage, and delivered across ByteDance's consumer apps and the Volcano Engine API, expected in August.