●Change8

v0.29.0

📦 comfyuiView on GitHub →
✨ 36 features🐛 4 fixes⚡ 1 deprecations🔧 50 symbols

Summary

This release introduces significant improvements in video processing, adds support for new AI models from various partners, and includes numerous optimizations and bug fixes across different components of ComfyUI.

Migration Steps

  1. Update workflow templates to v0.11.11.
  2. Update workflow templates to v0.11.12.
  3. Update workflow templates to v0.11.15.
  4. Update workflow templates to v0.11.17.
  5. Update workflow templates to v0.11.19.
  6. Update embedded docs to v0.5.9.

✨ New Features

  • Stream video transcodes instead of buffering every frame in RAM.
  • Send ComfyUI Job ID in request headers for Partner Nodes.
  • Add JoyImageEdit native model support.
  • Add GPT5.6 models for Partner Nodes (OpenAI).
  • Various optimizations and fixes for Comfy Kitchen.
  • Speed up Anima.
  • Support Anima Lite control models.
  • Add Gemini 3.5 Flash LLM model for Partner Nodes (Google).
  • Sync shared API contract from cloud.
  • Add Avatar, Talking Photo, Create Avatar, Video Translate, and TTS nodes for Partner Nodes (HeyGen).
  • More int8 optimizations for Comfy Kitchen.
  • Implement regular and timestep zero reference images to Krea 2 for Ostris and Identity Edit Ref LoRAs.
  • Support Gemma4 12B.
  • Video processing nodes, Image Processing Node video support, and trainer video support.
  • FreSca: 5D+ (ex. Anima) fix, model-agnostic iteration.
  • Add native Uni3C Controlnet support for Wan models.
  • Pass videos as inline data for Partner Nodes (Gemini-Omni).
  • Use optimized rms_rope function in JoyAI image model.
  • Add new models for Partner Nodes (OpenRouter).
  • Add new models for Partner Nodes (Anthropic).
  • Make z image/lumina 2 models use Comfy Kitchen rms rope.
  • Add new audio-1.0-multilingual model for Partner Nodes (ByteDance).
  • Make Ernie use Comfy Kitchen rms rope.
  • Use Comfy Kitchen rope functions in LTX models.
  • Convert latent_upsampler model to ModelPatcherDynamic.
  • Bump default background usage clamp to 128GB for CLI arguments.
  • Support MageFlow.
  • Improve LTXV IC-lora detection.
  • Add Claude Opus 5 for Partner Nodes (OpenRouter).
  • Optimize Ideogram model using Comfy Kitchen rms rope.
  • Fix MageFlow on cards that don't support bf16.
  • Add Claude Opus 5 model for Partner Nodes (Anthropic).
  • Allow using float fps for LTXVEmptyLatentAudio.
  • Update stable portable release workflow.
  • Respect "X-Comfy-Credits-Used" header from the comfy-api for Partner Nodes.
  • Add V4.1 models and V4 image edit nodes for Partner Nodes (Recraft).

🐛 Bug Fixes

  • Fix wan dancer issue with batches.
  • Fix gfx1035 not being treated like RDNA2.
  • Fix Gemini-Omni for Partner Nodes: pass videos as inline data.
  • Fix MageFlow on cards that don't support bf16.

Affected Symbols

VideoComfyUI Job IdJoyImageEditGPT5.6Comfy KitchenAnimaAnima LiteGemini 3.5 FlashOpenAIGoogleHeyGenAvatarTalking PhotoCreate AvatarVideo TranslateTTSKrea 2OstrisIdentity Edit Ref LoRAsWan dancerGemma4 12Bgfx1035RDNA2Uni3C ControlnetWan modelsGemini-Omnicursor-reviewgithub-workflowsJoyAIOpenRouterAnthropicz imagelumina 2Ernieltx modelslatent_upsamplerModelPatcherDynamiccli_argsMageFlowLTXV IC-loraIdeogramLTXVEmptyLatentAudiostable portable release workflowcomfy-apiRecraftClaude Opus 5ByteDanceaudio-1.0-multilingualRunwayGen3a

⚡ Deprecations

  • The Gen3a model from Runway is deprecated.