●Change8

v0.3.60

📦 comfyuiView on GitHub →
✨ 24 features🐛 6 fixes🔧 19 symbols

Summary

This release heavily focuses on integrating and refining support for the Hunyuan image model, alongside significant updates to API node schemas (V3 conversion) and the introduction of several new model supports across image, audio, and video generation.

Migration Steps

  1. If using API nodes, be aware that the default timeout for API nodes has been reduced from 7 days to 2 hours. Adjust workflows or configurations if longer timeouts are required.
  2. Users utilizing the qwen edit plus model should now use the new TextEncodeQwenImageEditPlus node.

✨ New Features

  • Support for hunyuan image distilled model.
  • Fast preview for hunyuan image.
  • Implement hunyuan image refiner model.
  • Add Output to V3 Combo type to match what is possible with V1.
  • Add noise augmentation to hunyuan image refiner.
  • API nodes: add kling-v2-1 model to the KlingStartEndFrame node.
  • Conversion of several API nodes (Minimax, WanCameraEmbedding, Cosmos, nodes_cond.py, CFG nodes, Canny, Moonvalley API nodes) to the V3 schema.
  • Enable Runtime Selection of Attention Functions.
  • Hunyuan refiner vae now works with tiled.
  • Support wav2vec base models.
  • Add support for Chroma Radiance.
  • Support the omnigen2 umo lora.
  • Add encoder part of whisper large v3 as an audio encoder model.
  • Support the 1.7B HuMo model.
  • Support the HuMo 17B model.
  • Enable fp8 ops by default on gfx1200.
  • Do padding of audio embed in model for humo for more flexibility.
  • Basic WIP support for the wan animate model.
  • Update WanAnimateToVideo to more easily extend videos.
  • Add inputs for character replacement to the WanAnimateToVideo node.
  • Add strength input to Differential Diffusion.
  • Support for qwen edit plus model, introducing the new TextEncodeQwenImageEditPlus node.
  • Add offset param to /history endpoint.
  • feat(api-nodes): add wan t2i, t2v, i2v nodes.

🐛 Bug Fixes

  • Fix hunyuan refiner blownout colors at noise aug less than 0.25.
  • Fix python -O and BatchedBrownianTree w0 handling.
  • Fix image sharpen node.
  • Fix LoRA Trainer bugs with FP8 models.
  • Fix bug with WanAnimateToVideo node.
  • Fix bug with WanAnimateToVideo.

🔧 Affected Symbols

KlingStartEndFrameMinimax API nodesWanCameraEmbedding nodeCosmos nodesnodes_cond.pyCFG nodesCanny nodeMoonvalley API nodesHunyuanImageBatchedBrownianTreeimage sharpen nodeLoRA TrainerWanAnimateToVideoDifferential DiffusionTextEncodeQwenImageEditPlus/history endpointwan t2i nodeswan t2v nodeswan i2v nodes