Multimodal Reference Conditioning of Image and Video Generation
the central 2026 capability shift — Seedance 2.5 takes up to 50 references (30 images, 10 video, 10 audio) and MiniMax H3 takes 9 images plus 3 video plus 3 audio, and production control now means assembling that reference set
This Concept is waiting for its first lesson!
the central 2026 capability shift — Seedance 2.5 takes up to 50 references (30 images, 10 video, 10 audio) and MiniMax H3 takes 9 images plus 3 video plus 3 audio, and production control now means assembling that reference set
Are you a teacher? Sign in to start contributing.
Sign In