ComfyUI Extension: ComfyUI-JoyAI-Image-Edit
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.
Native ComfyUI implementation of the JoyAI-Image family supporting single and multi-image image editing and text-to-image generation using JoyAI-Image models with Qwen3-VL support.
Looking for a different extension?
Custom Nodes (0)
README
This repository is the native ComfyUI implementation of the JoyAI-Image family, from the official Pull request submitted by JoyAI:
https://github.com/Comfy-Org/ComfyUI/pull/14428
Requirement: ComfyUI v0.25.0 or higher, which introduced native Qwen3-VL support (PR #14298).
Supported Models
| Model | Node | Official Repo |
|-------|------|---------------|
| JoyAI-Image-Edit (single-image edit / T2I) | TextEncodeJoyImageEdit | jdopensource/JoyAI-Image-Edit-ComfyUI |
| JoyAI-Image-Edit-Plus (multi-image edit / T2I) | TextEncodeJoyImageEditPlus (up to 4 reference images) | jdopensource/JoyAI-Image-Edit-Plus-Diffusers |
The Edit Plus model ships as diffusers shards. Merge them into a single safetensors with
convert_plus.py:cd custom_nodes/comfyui_joyimage_edit python convert_plus.py <downloaded_diffusers_folder> <output.safetensors>
Quantized Weights
NVFP4 model weights quantized with Comfy-kitchen:
https://huggingface.co/aha2023/JoyAI-Image-Edit-ComfyUI-NVfp4
Two quantization configs are included:
config_nvfp4.json— Balanced (attention FP8 + MLP NVFP4)config_nvfp4_extreme.json— Extreme (all layers NVFP4)
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.