TL;DR: quick fix "patch" LoRA to try and get better hands if stuff went wrong :D Will not do magic, but can help without breaking the original picture too much. Can be used as-is at base strength 1.0 or adapted between 0.8 and 1.2 if needed.Long story: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/11160/fixing-hands-is-such-a-pain-tt-part-1">Part 1</a> - <a target="_blank" rel="ugc" href="https://civarchive.com/articles/11180/fixing-hands-is-such-a-pain-tt-part-2">Part 2</a>I am releasing the 2 versions:<ul><li>2generate: 32 dimensions version that can be applied on the fly during generation</li><li>4merge: 128 dimensions version to be used for merge in checkpoint if needed</li></ul>NB: I have marked this model as Illustrious to allow usage in the Civitai Generator since the other "Hands" LoRA i found are marked either as Pony or SDXL1.0, but this should work with any SDXL based model :DUpdate: to help with Civitai Generator, i released the exact same file as 2generate tagged as a Pony LoRA. If you don't use the on-site generator, this "Pony" version is not needed if you already got the 2generate for Illustrious.

HandPatch (Unet-Only)

"Small" 32 dimensions version for on-the-fly usage

handPatchUnet.safetensors

2generate

2generate pony

4merge

This is the exact same file as the 2generate version, so if you are not using Civitai Generator, it is useless. But since the choice of Base Model lock what you can select there, here it is, as a "Pony" version.

Previously trained LoRa. Hand restoration, trained only on Uent, minimally impacting the image. Suitable for images with hand problems; note the seed and redraw the image. Not recommended for continuous use.0.5-1

Files with SHA256: 041ae75cf09970ed549b0217817af01b3b0e81baf8e0319d53b8c5e40d83ed8d