Anikin Labs — MiniMax P-H3-Image Workflow
A custom ComfyUI workflow for MiniMax P-H3-Image, designed to make high-quality image generation and reference-based image editing simple and accessible.
This workflow uses the PrunaAI W4A8 quantized version of P-H3-Image and combines two generation modes in one convenient interface.
Features
Text-to-Image: Generate detailed images from text prompts.
Reference Editing: Use up to two reference images to guide image generation and editing.
One-Click Mode Switching: Switch between Text-to-Image and Reference Editing using a simple toggle.
Adjustable Resolution: Experiment with image resolution settings.
Optimized Workflow Layout: Organized nodes for convenient everyday use.
Custom Sigma Schedule: Includes a predefined 8-step sampling schedule.
Required Model Files
Download the model files from:
https://huggingface.co/PrunaAI/P-H3-Image-Comfy
Required files:
Diffusion model
p-h3-image_w4a8.safetensors
Text encoder
p-h3-image_text_encoder_w4a8.safetensors
VAE
p-h3-image_vae_int8_convrot.safetensors
How to Use
Download and install the required model files.
Import the workflow JSON into ComfyUI.
Select your generation mode using the toggle.
Enter your prompt and, for Reference Editing, load your reference images.
Adjust the available settings and start generating!
Important: Keep the predefined sigma schedule unchanged unless you understand how to configure it for this model.
License Notice
This workflow does not include model weights. MiniMax P-H3-Image is subject to its own license terms, including geographic restrictions. Please review the official model license before downloading or using the model, particularly if you are located in the United States, European Union, United Kingdom, or South Korea.
Creator
Anikin Labs
Custom ComfyUI workflow by Roman Anikin.
If you find this workflow useful, please consider leaving a rating and sharing your results!