Exploring the Best Workflow for Face Recovery from Low-Res Photos

NetworkSpecial3268 · reddit · 2026-08-09

A Reddit user is seeking the optimal ComfyUI workflow to recover clear faces from extremely low-resolution photos (as small as 150x150 pixels) for use in LoRA training.

The user previously tried a FLUX.1 Dev Controlnet-based upscaler workflow, but it only produces grey noise in the new ComfyUI version. While testing new solutions like Adonis, they found that processing extremely small and grainy source images still causes artifacts like oversharpened eyes or dirty skin. They are asking for advice on pre-processing and post-processing techniques to overcome these limits.

Original post →

More from Multimodal

Multimodal channel →