v0.3.32
Comfy-Org/ComfyUIv0.3.32May 6, 2025by github-actions[bot]
AI Summary
This release adds more API nodes and a flag to disable them, while switching several model architectures to use native PyTorch RMSNorm for better performance.
Key Highlights
- Added `--disable-api-nodes` flag to control API node loading
- Switched Mochi, Wan, Cosmos, Hydit, and Lumina models to use native PyTorch RMSNorm
- Added support for saving Comfy VIDEO type to buffer
- Fixed error creating video when audio tensor chunks are non-c-contiguous
New Features
- More API nodes
- Disable API nodes via --disable-api-nodes flag
- VIDEO type saving to buffer
- Fix for video creation error
Full Release Notes
## What's Changed * More API nodes by @Kosinkadink in https://github.com/comfyanonymous/ComfyUI/pull/7956 * Add a way to disable api nodes: --disable-api-nodes by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/7960 * Switch mochi and wan modes to use pytorch RMSNorm. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/7925 * Change cosmos and hydit models to use the native RMSNorm. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/7934 * Change lumina to native RMSNorm. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/7935 * Support saving Comfy `VIDEO` type to buffer by @christian-byrne in https://github.com/comfyanonymous/ComfyUI/pull/7939 * Fix: Error creating video when sliced audio tensor chunks are non-c-contiguous by @christian-byrne in https://github.com/comfyanonymous/ComfyUI/pull/7942 * This should not be a warning. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/7946 **Full Changelog**: https://github.com/comfyanonymous/ComfyUI/compare/v0.3.31...v0.3.32