comfyui多人工作流 — v优化版

This ComfyUI text-to-image workflow creates images with multiple characters from character traits or tags and interaction actions.

Assign different actions to different characters, including characters whose positions overlap.

A special node affects conditioning sampling and changes model attention.

Add character traits or tags and the interaction actions you want to depict.

The output is a generated text-to-image scene showing multiple characters and their requested interactions.

Check the version information before use because functionality differs between versions.

Suggestion · not verified

The listed base model is Illustrious.

It requires these node packs: cg-use-everywhere, comfyui_densediffusion, efficiency-nodes-comfyui, and rgthree-comfy.

It also requires the miaomiaoHarem_v17.safetensors checkpoint.

For a lower-quality character LoRA, lower its weight to help avoid feature contamination.

Suggestion · not verified

When combining LoRA, tag-based, and trigger-word characters, add extra feature descriptions to the trigger-word character.

Suggestion · not verified

Three-character text-to-image scenes may be less stable because a higher-weight character can overwhelm a lower-weight character.

Suggestion · not verified