>>2777506> NAI devs made clip skip 2 a requirement of their model yes. they trained their model on clip skip 2. other than the loss of detail/accuracy, loras not working properly is the worst effect i've seen.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip>>2777482more comparisons:
https://files.catbox.moe/a2kz6x.jpghttps://files.catbox.moe/oagrnt.jpghttps://files.catbox.moe/cn3gg2.jpg