Seedance 2.5

Seedance 2.5 is an AI video generation model from ByteDance that creates up to 30‑second video clips from text prompts, image, audio, and video references in a single generation pass. It supports up to 50 multimodal inputs—roughly double the limit of its predecessor—and outputs native 4K video with 10‑bit color depth, preserving detail for downstream grading. The model includes local re‑draw editing, letting users swap a product, background, or character while keeping motion, camera path, and lighting intact, which is useful for e‑commerce product variants, marketing drafts, and agency client reviews. Targeted at creators, marketers, agencies, and e‑commerce teams, Seedance 2.5 aims to deliver longer, reference‑heavy video drafts that require fewer regeneration cycles and higher prompt adherence.