Spent the last few days trying to optimize AI image generation on ComfyUI. Installed the GGUF model, LoRA, VAE, and text encoders, organized them into the right folders, and finally got the Lightning v1.1 LoRA working.

Added Sage Attention + FP16 Accumulation to reduce VRAM usage and speed things up. Configured the CFG to 4.5 — and wow, images are razor-sharp, vibrant, and super clean compared to v1.0 or LightX2V.

If you’re into AI art or just want faster, high-quality results on 8GB VRAM, I highly recommend trying this workflow. Full tutorial On My Youtube Channel