How to use one reference image and a sketch in ComfyUI for character-consistent headshots

Big_CokeBelly · reddit · 2026-07-28

The post asks how to generate a headshot of a character in a different art style and expression using only one reference image, a hand-drawn pose/expression sketch, and ComfyUI.

The author says they cannot train a LoRA from a single commissioned image, but want to feed a character sketch plus pose and expression constraints into a node workflow for text-to-image generation.

Original post →

More from Multimodal

Multimodal channel →