🎬 ByteDance Announces Seedance 2.5 Video Model
At the Volcano Engine FORCE 2026 conference, ByteDance unveiled the Seedance 2.5 video model, which generates up to 30 seconds of continuous video in a single segment. The model supports up to 50 multimodal references for style control. Additionally, Doubao 2.1 Pro for coding and agentic tasks, and Seedream 5.0 Pro were also presented.
🌍 The transition to 30-second scenes increases video coherence and reduces editing costs, bringing AI closer to professional directing tools.
👤 ML researchers should take note of the capabilities for native long-segment generation without stitching artifacts.
Source 1: https://news.aibase.com/news/29085 Source 2: https://pandaily.com/bytedance-doubao-2-1-seedance-2-5-jun2026