ByteDance Launches Seedance 2.5: Generates 30s Videos, Accepts 50+ Multimodal Inputs
赛博禅心 · wechat · 2026-07-31
ByteDance's Seed team has officially released Seedance 2.5, its latest AI video generation model. The upgrade focuses on long-form storytelling and multimodal capabilities, extending single-shot high-quality video generation to 30 seconds with support for multi-round extensions.
Key upgrades include:
- Extensive Multimodal References: Accepts up to 30 images, 10 videos, and 10 audio clips simultaneously to accurately reproduce complex group scenes and multi-camera setups.
- Precise Timestamp Editing: Allows targeted modifications of characters, actions, and audio within specific timeframes, featuring advanced green screen replacement and camera movement adjustments.
- Visual & Physics Refinement: Systematically reduces the unnatural "AI gloss" and improves the realism of lighting and physical rules.
The model is now available on Jimeng AI and Doubao Pro, with API access coming to Volcano Engine soon. It can also generate synthetic data for industrial manufacturing, embodied AI, and autonomous driving training.
Related event: ByteDance Releases Seedance 2.5: Native 30s Video and Multimodal Control(60 posts)→
More from Multimodal
- Alibaba Opensources ClinFusion: A Medical Multimodal Foundation Model — aigclink · 2026-08-01
- MiniMax 3 passes Turing test for AI video: writes 'Hi' on chalkboard, now open source — Kyrannio · 2026-08-01
- Testing Doubao vs. Jimeng: $30 Gets You Only 3-6 AI Video Generations — oran_ge · 2026-08-01
- Experiment Shows Training AI Image Models at 512 Resolution + Upscaling Saves VRAM — More_Bid_2197 · 2026-08-01
- 8-Hour Debug of ComfyUI Black Images Uncovers PyTorch FP16 Overflow Bug — rtsitola · 2026-08-01
- MoGe-3 by Microsoft Sets SOTA on 9 Benchmarks for High-Fidelity 3D Geometry from a Single Image — RexDouglass · 2026-08-01