Seedance 2.5 generates 30-second consistent videos with massive asset reference
thetripathi58 · x · 2026-08-24
Seedance 2.5 is now live inside Magnific, allowing users to direct actual scenes with AI. The tool supports generating up to 30-second continuous takes without stitching. It handles complex interactions, such as multiple characters laughing simultaneously, without breaking consistency. The key differentiator is a massive reference system that accepts up to 50 assets per generation (30 images, 10 videos, 10 audio tracks). Users can control camera zoom, focus, and depth of field, and retouch specific details without regenerating the whole video.
More from Multimodal
- Ostris AI Toolkit ships day-0 training support for FLUX.3 Image Klein — robrombach · 2026-10-16
- LTX partners with Asteria to bring open-weight video models into pro filmmaking — makeitrad1 · 2026-10-16
- Kling 4.0 now fully available, showcased with 'King Card' short film — DavidmComfort · 2026-10-15
- Krea 3 image editing beta opens: 2k edits in under 8s, claims Nano Banana Pro-level quality — TitusTeatus · 2026-10-15
- Diffusers adds full Flux.3 Image Klein support: FP8/NVFP4 quantized inference and bbox layout control — RisingSayak · 2026-10-15
- xiaohu demos video model across styles with in-video text rendering — xiaohu · 2026-10-15