Description
Image imported from Derpibooru.
Original: https://derpibooru.org/images/3168659
Metadata at the time of import:
Original: https://derpibooru.org/images/3168659
Metadata at the time of import:
| Field | Data |
|---|---|
| Favorites | 122 |
| Upvotes | 162 |
| Downvotes | 25 |
| Score | 137 |
| Comments | 8 |
| Uploader | aiLex |
Original description below this line
These are the most accurate feet I have ever been able to get from AI.
Unfortunately, AI is awful in feet generation, especially when they are as big as half of a picture. It took literally thousands of generations and a lot of inpainting afterwards. And result still has a lot of failures (like the left and right feet are mixed up). But at least her feet are really look like… um… feet.
Let’s wait for PonySDXL release, it should do such tasks much more successfully.
Unfortunately, AI is awful in feet generation, especially when they are as big as half of a picture. It took literally thousands of generations and a lot of inpainting afterwards. And result still has a lot of failures (like the left and right feet are mixed up). But at least her feet are really look like… um… feet.
Let’s wait for PonySDXL release, it should do such tasks much more successfully.
What would you like to know?
SDXL is the latest image generation model published recently. It’s much better in terms of composition and allows you to create high resolution images with less VRAM consumption. Support for this model has already been added to the popular WebUI framework from AUTOMATIC1111 and several other tools. Unfortunately, at the moment it is not adapted for the generation of MLP images.
But AstraliteHeart, trainer of well-known Pony Diffusion models, said he would eventually train a pony version of this model. It will be much better than current Pony Diffusion.
Yeah, they look like they are molded from sweet cream chocolate.