JoyAI Image Edit: Native ComfyUI Support for Image Editing

JD Open Source has released native ComfyUI support for JoyAI-Image-Edit and JoyAI-Image-Edit-Plus, bringing their powerful instruction-guided image editing models directly into the ComfyUI ecosystem.

What is JoyAI-Image?

JoyAI-Image is a unified multimodal foundation model developed by JD Open Source (JD.com). It combines an 8B Multimodal Large Language Model (MLLM) with a 16B Multimodal Diffusion Transformer (MMDiT) to deliver image understanding, text-to-image generation, and instruction-guided image editing in a single architecture.

The model family is built around a closed-loop collaboration between understanding, generation, and editing — stronger spatial understanding improves grounded generation, while generative transformations provide complementary evidence for spatial reasoning.

ComfyUI Support

As of June 9, 2026, both JoyAI-Image-Edit and JoyAI-Image-Edit-Plus run natively in ComfyUI without any additional dependencies. The ComfyUI-optimized model repositories provide pre-packaged model files in two formats:

• BF16 — Full precision for maximum quality
• INT8 ConvRot — Quantized for lower VRAM usage with minimal quality loss

A bundled ComfyUI workflow is included in each repository for immediate use.

JoyAI-Image-Edit

Single-image instruction-guided editing. Accepts an input image and a natural language instruction (e.g., "Turn the plate blue" or "Move the cup to the right") and generates the edited result.

• HF Repo: jdopensource/JoyAI-Image-Edit-ComfyUI
• Model Files:
  • diffusion_models/joyai_image_edit_bf16.safetensors
  • diffusion_models/joyai_image_edit_int8_convrot.safetensors
  • text_encoders/qwen3vl_8b_joyimage_edit_bf16.safetensors
  • text_encoders/qwen3vl_8b_joyimage_edit_int8_convrot.safetensors
  • vae/wan_2.1_vae.safetensors

JoyAI-Image-Edit-Plus

Multi-image instruction-guided editing. Accepts 1 to 6 reference images along with a text instruction, and generates a new image combining elements from all references according to the instruction. Supports multi-image composition, consistency, and joint manipulation.

• HF Repo: jdopensource/JoyAI-Image-Edit-Plus-ComfyUI
• Model Files:
  • diffusion_models/joyai_image_edit_plus_bf16.safetensors
  • diffusion_models/joyai_image_edit_plus_int8_convrot.safetensors
  • text_encoders/qwen3vl_8b_joyimage_edit_plus_bf16.safetensors
  • text_encoders/qwen3vl_8b_joyimage_edit_plus_int8_convrot.safetensors
  • vae/wan_2.1_vae.safetensors

Installation

Place the downloaded files in the following ComfyUI directory structure:

ComfyUI/
└── models/
    ├── diffusion_models/
    │   ├── joyai_image_edit_bf16.safetensors
    │   └── joyai_image_edit_int8_convrot.safetensors
    ├── text_encoders/
    │   ├── qwen3vl_8b_joyimage_edit_bf16.safetensors
    │   └── qwen3vl_8b_joyimage_edit_int8_convrot.safetensors
    └── vae/
        └── wan_2.1_vae.safetensors

Alternatively, use hf download to install directly:

hf download jdopensource/JoyAI-Image-Edit-ComfyUI --local-dir /path/to/ComfyUI/models

Capabilities

JoyAI-Image-Edit excels in several areas of instruction-guided editing:

• Spatial Editing — Object move, object rotation, and camera control through natural language prompts
• Precise Instruction Following — High-fidelity adherence to editing instructions while preserving unmodified scene elements
• Multi-Image Composition (Edit-Plus) — Combine elements from up to 6 reference images into a single cohesive output
• Strong Scene Preservation — Maintains scene structure and content outside the edited region

The model's spatial intelligence is also leveraged for spatial reasoning tasks — it can synthesize diagnostic viewpoints by faithfully executing camera motions, helping disambiguate complex spatial relations.

Links