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.
More from Multimodal
- BFL Releases FLUX 3 Video Model, Open Weights Confirmed — cloneofsimo · 2026-08-05
- ByteDance Integrates Seedance Video Model into Gauth for EdTech Animations — Economy_Cicada8756 · 2026-08-05
- Testing Open-Source AI Workstation Moonlit: End-to-End Picture Book Generation — sujingshen · 2026-08-05
- Generating 10-Minute AI 'Seinfeld' Episode Using Minimax h3 and GLM 5.2 — nathandreamfast · 2026-08-05
- MiniMax Video Model Demo: Generating Tom Holland Eating Street Food from 4 Images — CurieuxExplorer · 2026-08-05
- MiniMax H3 Early Access Live: Native Multimodal Generation & Precise Editing — HeyAmit_ · 2026-08-05