Testing Krea2 Face Swap LoRA: ComfyUI Workflow & Optimization Tips

cradledust · reddit · 2026-08-05

A Reddit user shared their experience testing the Krea2BFSv1 LoRA and ComfyUI workflow for AI face swaps. The workflow takes about 1.4 to 2 minutes per generation on an 8GB RTX 4060. The author noted that tweaking the Fidelity Dial allows for full-body or head-only swaps, emphasizing that finding the right source image for maximum likeness is an art.

Original post →

More from Multimodal

Multimodal channel →