Recommended Weight: 0.6~1.0
Trigger Words:
lingsha, hair ornament, red limbs, single wrist cuff, leg tattoo, thigh strap, black heels
Description
Trained based on Pony V6 /w the updated dataset.
Revised trigger words for better consistency.
FAQ
Comments (4)
v1 tends to deep fry my image when I put CFG to 5 and sampling steps to 20, but when I put it to 7/30 as shown in the showcase images, it works as fine. It never did this in the alpha.
Overall, great LoRA.
Thanks for the feedback! Would you mind telling me which checkpoint (and prompts) were used when the issue occurred? For the sample images, 7/30 was used because I upscaled them in img2img. For text2img, I usually use 7/25 or 6/25.
@fallenL I used autismmix confetti to generate the images. Here's the PNG info from the image that caused the issue: https://i.imgur.com/FEUsnmC.png
@ipsylon0000 Hmm, interesting. I tried using the exact same prompts, checkpoint, VAE, and the same other generation settings (seed, CFG, steps, etc.) except for ADetailer (which I believe shouldn't be the cause of the problem), but I wasn't able to replicate your result. Here's my result: https://imgur.com/a/lingsha-test-CYuSxXw
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.