Description
Image imported from Derpibooru.
Original: https://derpibooru.org/images/2969427
Metadata at the time of import:
Original: https://derpibooru.org/images/2969427
Metadata at the time of import:
| Field | Data |
|---|---|
| Favorites | 51 |
| Upvotes | 83 |
| Downvotes | 7 |
| Score | 76 |
| Comments | 3 |
| Uploader | Siber |
Original description below this line
An AI take on a fandom classic!
Generation parameters
lyra heartstrings sitting on a bench like a human, pony, unicorn, solo, digital art
Negative prompt: bad anatomy, extra limbs, extra horns, oversaturated colors
Steps: 40, Sampler: Euler a, CFG scale: 9, Seed: 2323570357, Size: 512x512, Model hash: 9070b574, Batch size: 8, Batch pos: 2, Denoising strength: 0.5, Mask blur: 4
lol. Thanks for the reply. If it works, it works. Maybe I’ll try “…,artist shakes rubber chicken and chants,…” as well when I keep getting and abada instead of a unicorn. Like you said, for now, inpaint, patience, gimp go a long way.
It definitely affects the output. Whether in a good or a bad way - that is a big question. It’s all, basically, statistical alchemy anyway - if you believe it works, then it works :-) Also do keep in mind that this isn’t the raw output. I had to use the “inpaint” tool (allows you to “fix” parts of the image using the same neural network) a lot on it to make it look somewhat presentable. Another annoying tendency is ponies pretty much always come out with wrong eye color. Currently I was unable to come up with a recipe to fix that, other then finally giving up and manually correcting it in GIMP :-)