As always I'm not quite happy with the results, but I haven't posted in a while, so here it is.

Thanks to Aktiosa for providing the dataset)

Tags

• Feet pose

• Soles

Trained resultions

• 768x512

• 768x768

• 512x768

*Haven't tried on other resolutions

Params

Kinda picky on sampler, so I recomend using one of the folowing:

• dpm++ 2M SDE Karras

*"euler a" provides okeish results too, but it's not that good in img2img in my opinion

And for the other parametrs

• Clip skip - 2

• Lora weight - 0.6 worked the best for me (photorealistic models get overburned if add more weight, but you can fix it in img2img)

Also I suggest using txt2img with 0.6 weight, and then use img2img with lower (0.5, or less) weight of the Lora. Delievers a decent results for me.

Plans

Working on stablility and variebility of the model. earlier version were much more unstable.

Enjoy)

Feel free to ask any questions here in comments, or in my discord channel.

Also I am making games with AI arts. They will be free in the future, but if you want to participate in making or have an early access for Loras I am currently working on - you can )

Also my wet hair Lora was banned here, I dunno why. You can the last version of it from my patreon for free.