Type | |
Stats | 906 4,824 |
Reviews | (97) |
Published | Apr 18, 2024 |
Base Model | |
Training | Steps: 25,000 Epochs: 20 |
Usage Tips | Clip Skip: 1 Strength: 0.8 |
Trigger Words | sts-foreshortening-v2 |
Hash | AutoV2 DED44C91E2 |
This LoRA is for PonyDiffusion v6 XL - for best results use PonyDiffusion v6 XL or merges based off of PonyDiffusion v6 XL like AutismMix.👀👇READ ME 👇👀
(no seriously, read the description below)
[Updated for v2.0]:
This LoRA (technically LyCORIS) is for adding foreshortening to your images. There are two v2.0 versions - one LyCORIS and one LoKR. I recommend you use the LyCORIS, as the LoKR was made to have a super small and compressed version. As a result, the LoKR version is not as flexible.
Activation tag for v2.0 LoRas is sts-foreshortening-v2 . Recommended weight 0.6 to 1.0 . Because of the sheer percentage of foreshortening pictures on the internet being foot/feet related, this LyCORIS does foot_focus, feet, and soles rather well. This LyCORIS also does arms_outstreched and reaching_towards_viewer well. This LyCORIS pairs well with my detail slider LoRA for PonyXL.
For examples prompts, refer to the included example images. All the metadata is correct in the them except that they were generated with AutismMix Confetti and not SeraphMix4.
Changelog:
Feb. 27 2024 - Initial Release w/ UNET only version
April 18 2024 - Released v2.0 LyCORIS and LoKR versions
~~~
OLD DESCRIPTION FOR V1.0:
This LoRA (technically LyCORIS) is for adding foreshortening to your images. The main thing to know is the difference between the UNET Only version and the standard version. The UNET version stays closer to what would be generated without this LyCORIS, so things like pose, background, and general detail from the original checkpoint can be preserved better. The standard version is trained on both UNET and TE, so it is more transformative. Neither one is inherently better from my testing.
Activation tag for 1.0 LoRAs is foreshortening. Because of the sheer percentage of foreshortening pictures on the internet being foot/feet related, this LyCORIS does foot_focus, feet, and soles rather well. I did remove the foot_focus tag from the training tags to try and make sure this didn't end up being a glorified foot_focus LyCORIS. This LyCORIS also does arms_outstreched and reaching_towards_viewer well.