🎬 ByteDance has released Seedance 2.5 — a new video generation model that creates continuous 30-second clips in 4K resolution in a single pass.
Thanks to its Sparse Diffusion Transformer architecture, the model solves the problem of character and lighting "drift" by supporting up to 50 multimodal references for maximum control.
🌍 The shift to full 30-second scenes radically changes video production by minimizing manual stitching. This positions ByteDance as a leader in temporal coherence.
👤 It is now possible to create complex and long clips while maintaining character appearance and environmental stability through a single prompt.
Source 1: https://aivideoadvisor.com/seedance-2-5-native-30-second-4k-ai-video-generation-launches-july-2026/ Source 2: https://www.techtimes.com/articles/320683/20260716/seedance-25-api-live-bytedances-30-second-ai-video-carries-unresolved-copyright-risk.htm