ByteDance's Seedance 2.5 generates 30-second videos from 50 inputs

ByteDance has officially launched Seedance 2.5, a significant leap in AI-driven video generation. The model can produce high-quality 30-second video clips in a single inference pass, setting a new benchmark for single-shot video duration among rival systems. This capability directly addresses a common limitation in the field, where most models generate only 5-10 second segments that require stitching together. By enabling longer coherent outputs, Seedance 2.5 reduces post-production work and allows for more complex storytelling.

Seedance 2.5 is being integrated directly into ByteDance's consumer products, Jimeng AI and Doubao Pro, making the technology accessible to a broad user base. Developers will also gain programmatic access through the Volcano Ark platform, which provides API endpoints for building custom applications. A standout technical feature is support for up to 50 multimodal reference inputs, meaning users can feed the model multiple images, text prompts, and other media types simultaneously to guide the generation with exceptional precision. This positions Seedance 2.5 as a versatile tool for creators, marketers, and developers seeking to produce long-form AI videos with granular control over character, style, and scene continuity.