●Change8

v0.3.35

📦 comfyuiView on GitHub →
✨ 14 features🐛 7 fixes🔧 9 symbols

Summary

This release introduces significant feature enhancements, including new audio output formats (opus, mp3), string operation nodes, APG guidance implementation, and improved API client robustness. Several bug fixes address issues related to RMSNorm, APG momentum, and API request retries.

Migration Steps

  1. If you rely on the default delimiter for the concatenate node, you may need to explicitly set it now that the default has been removed.

✨ New Features

  • Added support for opus and mp3 audio output formats in the audio output node.
  • Added nodes for basic string operations.
  • Added support to record video as output for the 3D node.
  • Implemented APG guidance.
  • Added comment in example script detailing how to use API keys without needing the frontend.
  • Added support for official ACE Step loras.
  • Hunyuan Custom initial untested implementation added.
  • Improved robustness of client.py and added logging of API requests.
  • Display progress and result URL directly on API nodes.
  • Added `get_duration` method to Comfy VIDEO type.
  • Added Wan-FUN Camera Control models and the WanCameraImageToVideo node.
  • ImagePadForOutpaint now returns a 3 channel mask.
  • Added a node to add pixel space noise to an image.
  • Added delimiter option to the concatenate node.

🐛 Bug Fixes

  • Fixed issue with old pytorch RMSNorm calculation.
  • Fixed negative momentum calculation in the APG node.
  • Fixed bug where the API client would retry a request even if the API reported a task failure.
  • Fixed typo in a previous PR.
  • Fixed missing category for T5TokenizerOption.
  • Added validation for video inputs.
  • Removed default delimiter from the concatenate node (implying a change in default behavior or removal of a previous default).

🔧 Affected Symbols

audio output node3d nodeclient.pyAPI nodesComfy VIDEO typeWanCameraImageToVideo nodeImagePadForOutpaintT5TokenizerOptionconcatenate node