ComfyUI v0.31.0

v0.31.0
Added 7
  • Add Bloom 2 and Wonder 3.5 models support for TopazAI
  • Add Flux 3 video model support for BFL
  • Support int8_convrot VAE for MiniMax-H3
  • Support Wan-Animate2
  • Support asym w4a8_int
  • Add SeeDance 2.5 model support for ByteDance
  • Add ImageCompositor node with layer-state compositing and Seedream Layer Separation node
Changed 2
  • Avoid pinning too much memory on Linux systems with no swap partition
  • Speed up LTX and Wan
Fixed 8
  • Cast raw parameters to input device in MiniMax H3 VAEs
  • Harden some nodes against potential issues related to combos
  • Restore SDPA non-cudnn small attention bypass
  • Fix MiniMax H3 noise mask sampling
  • Fix sampler issues for audio with minimax and support more samplers
  • Skip creating useless noise in res_multistep sampler
  • Fix full offload on minimax audio VAE
  • Fix MiniMax H3 audio corruption with EasyCache
Deprecated 1
  • Deprecate Reve nodes
What's Changed
New Contributors

Full Changelog: https://github.com/Comfy-Org/ComfyUI/compare/v0.30.0...v0.31.0

View original

Upgraded? How did it go?

Discussion