I'm excited to release another concept Lora since the others had really great interaction : Split Pose
Recommended weight 0.6 - 0.8 ( if you increased you will get a lot of noise )
( you need to be specific in your prompts and might need few generation to land the pose you want since the pose is super complex)
Works perfectly with character and style Loras
For further support, feel free to DM me on Discord @.gamakichi_
Happy creating with Split Pose !
Description
clip skip 1-2
weight 0.5-0.8
FAQ
Comments (12)
Incredible 😮
Greetings !
Just want to let you know the concept have been mostly trained on Square and Landscape and Panorama Images so try avoiding vertical images as much as possible to ensure the Lora work as intended if you still have more questions feel free to ask here or DM ( btw it took me some retries to generate the showcase images as I like them to be, so if it don't generate as intended in the first time its completely normal) !
to force split you can either add more prompts related to splittings legs or increase the weight of the lora (0.8-0.9) is a good spot. Also look at the image prompts in the showcase for guidance
just make sure you generated like 4-5 images on the weight since it really varies if no split occured then do what I said in the previous comment
thank you for this ! Looks like a finally working great split lora :)
awesome <3 which resolution you used? 1024 x 1024?
yes mostly of the showcase images are 1024*1024 with an upscale of 2.
I made other images with (1152*896) / (1216 * 832) and turned out really great too
certain images require certain resolution so it might take some time to tweak the desired prompt
@Ayanna_01 Thank you so much <3 means a lot :)
@Ayanna_01 There is also a concept of crossed arms but they're not tucked in, like Nico Robin Pose from One Piece, it's also very good, hopefully someday we see it :), thank you for amazing work
Any chance of retraining this for Illustrious?
sure it will mostly take a week or so since I have a couple of things on list
@Ayanna_01 If you have a dataset, I can give it a shot at training the Lora. Also send me whatever settings you used for this one, if you have them. I'll send you back the file so you can update this page.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.