MOSS-Image LoRA Test ComfyUI Workflow (for LoRAs trained with KohakuMoss)
=========================================================================

Article: https://freakyjolly.com/kohakumoss-moss-image-lora-training/

What's inside:
  - MOSS_Image_LoRA_Test.json: ComfyUI workflow (UI format). Drag it onto the ComfyUI canvas or use Workflow > Open.

Required models (download each file and put it in the folder shown):
  - flux2-vae_small.safetensors
      folder: ComfyUI/models/vae/
      download: https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/vae/flux2-vae_small.safetensors
  - moss_image_sft_int8_convrot.safetensors
      folder: ComfyUI/models/diffusion_models/
      download: https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_int8_convrot.safetensors
  - qwen3.5_9b_w4a8_gptq.safetensors
      folder: ComfyUI/models/text_encoders/
      download: https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/text_encoders/qwen3.5_9b_w4a8_gptq.safetensors
  - your_moss_image_lora.safetensors
      folder: ComfyUI/models/loras/
      download: your own LoRA exported from KohakuMoss (placeholder name, select your file)

Custom nodes:
  - None. Every node is part of ComfyUI core (update ComfyUI to the latest version).

How to use:
  1. Update ComfyUI to the latest development build and restart.
  2. Download the three MOSS-Image files into the folders listed.
  3. Copy your exported KohakuMoss LoRA into ComfyUI/models/loras/.
  4. Drag MOSS_Image_LoRA_Test.json onto the canvas, open the subgraph, pick your file in "Your KohakuMoss LoRA" (strength 1.0, lower it if the effect is too strong), put your trigger word at the start of the prompt and queue.

Notes:
  - your_moss_image_lora.safetensors is a placeholder name: select your own LoRA file in the node.
  - The LoRA sits between the MOSS-Image model loader and the sampler, as described in the article. Test with this standard 25 to 50 step workflow before combining with a speed-up adapter.

Credits and licence:
  - Built on the official Comfy-Org MOSS-Image text-to-image template (MIT licence); the only change is one core LoraLoaderModelOnly node after the model loader.

Model weights are not included. Each model keeps its own licence; check the model page before commercial use.
Workflow shared by FreakyJolly.com.
