ZetaLvX MiniMax H3 Image – ComfyUI Workflow

A streamlined ComfyUI workflow for running MiniMax H3 Image locally, supporting text-to-image generation, image editing, and multi-reference generation.

The workflow includes adjustable aspect ratios, automatic resolution calculation, and direct control over the target megapixel count.

Youtube video - demonstration

Features

• Text-to-image generation

• Image editing using source images

• Multi-reference image conditioning

• Up to 16 image inputs

• Adjustable aspect ratios and resolution

• Direct generation without additional upscaling

• Organized nodes with built-in usage notes

• No custom nodes required

Required Models

Diffusion model
minimax_h3_image_pruned_int8_convrot.safetensors

Text encoder
qwen3vl_32b_minimax_h3_int8_convrot.safetensors

VAE
minimax_h3_image_vae_fp16.safetensors

Recommended Settings

• Steps: 6

• Sampler: Euler

• Scheduler: Simple

• CFG: 1

• Denoise: 1

• Target resolution: adjustable

How to Use

Text-to-Image: Leave the image inputs disconnected and enter your generation prompt.

Image Editing: Connect a source image to the first image input and describe the changes you want.

Multi-Reference: Connect multiple images and describe how they should be used in the prompt.

MiniMax H3 Image refers to connected images as <Picture 1>, <Picture 2>, and so on, following their input order.

The workflow contains usage notes and model information for easier setup.

Links

YouTube: https://youtube.com/@zetalvx

GitHub: https://github.com/ZETALVX

Patreon: https://www.patreon.com/Zetalvx