Starstruk Logo MarkStarstruk

v0.3.58

Comfy-Org/ComfyUIv0.3.58Sep 10, 2025by github-actions[bot]

AI Summary

This release adds support for Hunyuan 3D 2.1 and addresses several AMD-specific optimizations, including bf16 VAE support. It also introduces StabilityAI Audio API nodes.

Key Highlights

  • Support for Hunyuan 3D 2.1 model
  • AMD optimizations: Enable bf16 VAE on RDNA4 and fix lowvram issues
  • StabilityAI Audio API nodes addition
  • Support for Hunyuan image 2.1 regular model

New Features

  • Hunyuan 3D 2.1 Support
  • StabilityAI Audio API nodes
  • Qwen inpaint controlnet support
  • Fix progress update crossover between users

Full Release Notes

## What's Changed
* Fix progress update crossover between users by @guill in https://github.com/comfyanonymous/ComfyUI/pull/9706
* Add Hunyuan 3D 2.1 Support by @yousef-rafat in https://github.com/comfyanonymous/ComfyUI/pull/8714
* Some changes to the previous hunyuan PR. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9725
* fix: add cache headers for images  by @arjansingh in https://github.com/comfyanonymous/ComfyUI/pull/9560
* Fix lowvram issues with hunyuan3d 2.1 by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9735
* Print all fast options in --help by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9737
* Enable bf16 VAE on RDNA4. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9746
* Don't enable pytorch attention on AMD if triton isn't available. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9747
* Fix amd_min_version crash when cpu device. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9754
* Add ZeroDivisionError catch for EasyCache logging statement by @Kosinkadink in https://github.com/comfyanonymous/ComfyUI/pull/9768
* Add explicit casting in apply_rope for Qwen VL by @contentis in https://github.com/comfyanonymous/ComfyUI/pull/9759
* Support qwen inpaint controlnet. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9772
* API nodes: add ByteDance video by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9712
* Small refactor of some vae code. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9787
* Change validate_inputs' output typehint to 'bool | str' and update docstrings by @Kosinkadink in https://github.com/comfyanonymous/ComfyUI/pull/9786
* Fix issue on old torch. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9791
* Support hunyuan image 2.1 regular model.  by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9792
* Update template to 0.1.76 by @comfyui-wiki in https://github.com/comfyanonymous/ComfyUI/pull/9793
* Fix lowvram issue with hunyuan image vae. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9794
* API nodes: add StabilityAI Audio API nodes by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9749

## New Contributors
* @yousef-rafat made their first contribution in https://github.com/comfyanonymous/ComfyUI/pull/8714
* @arjansingh made their first contribution in https://github.com/comfyanonymous/ComfyUI/pull/9560

**Full Changelog**: https://github.com/comfyanonymous/ComfyUI/compare/v0.3.57...v0.3.58