READ ME!
The ideal weighting for this model is roughly 0.8
to 0.4
. Higher weighting gives better outfit accuracy, but less flexible in poses and background. You can also try decoupling the LoRA Unet and TE weighting with <Lora:UNET:TE> for max versatility.
Unfortunately there aren’t a ton of images of FL Chan, so I had to make do with a little under thirty images. I did manually process a number of them to make them as best as they could be, so hopefully that’s evident in the quality of this LoRA 🙂
Example prompt: stsflchan, 1girl, (singing, holding microphone), sitting, full body, green hair, ahoge, short hair, elbow gloves, headphones, red eyes, [fingerless gloves, skirt, knee boots], looking at viewer, bare shoulders
Example Chibi prompt: stsflchan, 1girl, (chibi), full body, green hair, ahoge, short hair, elbow gloves, headphones, red eyes, [fingerless gloves, skirt, knee boots], looking at viewer, bare shoulders, skindentation, toon \(style\)
As always, I’m always open to feedback on how to make better LoRAs, or LoRA requests.
Changelog:
November 22, 2023: Initial model release