Something I'm surprised I haven't seen yet, this should allow whatever character(s) you're generating to truly become bnuuy. Perfect hands not included.
For general usage, I'd recommend a weight of 0.8, but you can raise or lower it without much fuss, as well as the tags arms up and/or rabbit pose.
Arms up will work on it's own, but rabbit pose should be very effective in getting the pose. You can also try to use one arm up, best paired with something for the other hand to do, but I can't guarantee this one will work. It might depend on your checkpoint.
Additionally, if you're not using a character lora (or that lora isn't that strong when it comes to the outfit) you might want to put bunnysuit and ????? ears in the negative prompt, or it could end up altering the outfit.
Due to the common nature of images with the pose, this lora is very effectively trained on bunnysuits, ????? ears, and from front. You can avoid the outfit and ears, like I said above, but the angles are a different matter - in some situations, from [the] side can produce weird results, and it's more likely if the generation isn't trying to point the character's face and hands at the viewer.
??????????? has issues with the palms, as it will attempt to face the hands towards the camera. You can inpaint the hands, or use a 2nd pass (potentially with a lower weight) if you're using comfyui, in order to fix them however.
As this was trained on drawn images, I can't guarantee this will work with realistic models, but it should work well with most loras and checkpoints, so long as they don't have an iron rule over the character's arms.
Compared to the (unreleased) v1, this lacks images that obscure the hands, which should hopefully help with better looking hands.
1. I diritti dei modelli ripubblicati appartengono ai creatori originali.
2. I creatori originali che desiderano reclamare il proprio modello devono contattare il personale di SeaArt AI tramite canali ufficiali. Clicca per rivendicare
