When you say img2img do you mean you can give the AI a picture from reference?
I never used autismmix(better than purplesmart?), but one of the things I wanted to do with AI is take a picture and apply an artists style to it.
Yeah, you can also choose how closely the result should resemble the initial image. You can use the best result as a base for the next img2img and repeat x times to refine the picture.
I consider AutismMix and its derivatives to be overall better than Pony Diffusion V6 (used in PurpleSmart.ai), even if somewhat rigid.
Sure, you can use img2img with artist’s LoRA to get this effect, but I’m not sure how it would look, I haven’t experimented with artist LoRAs yet.
That certainly sounds superior. When I try to reroll an image, it’s always vastly different, when I just want to change a certain factor (like Applejack having her hat or just adjusting a face that is derped).
It’s also annoying when trying to certain ponies to be on top of each other with PurpleSmart. It will always put Rainbow Dash on top of Applejack or Rarity on top of Rainbow Dash no matter how I prompt it.
When you say img2img do you mean you can give the AI a picture from reference?
I never used autismmix(better than purplesmart?), but one of the things I wanted to do with AI is take a picture and apply an artists style to it.
Yeah, you can also choose how closely the result should resemble the initial image. You can use the best result as a base for the next img2img and repeat x times to refine the picture.
I consider AutismMix and its derivatives to be overall better than Pony Diffusion V6 (used in PurpleSmart.ai), even if somewhat rigid.
Sure, you can use img2img with artist’s LoRA to get this effect, but I’m not sure how it would look, I haven’t experimented with artist LoRAs yet.
The prompts look simular to the ones I use with Purplesmart AI.
When you say img2img do you mean you can give the AI a picture from reference?
I never used autismmix(better than purplesmart?), but one of the things I wanted to do with AI is take a picture and apply an artists style to it.
It was something like below, although the picture took some img2img and manual corrections to get right. I think my prompts are pretty bad overall, but I generate a lot until I stumble upon something decent and use it as a base for iterative refinement.
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, detailed, beautiful, masterpiece, best_quality, (duo), barn, barrel, detailed_background, close-up, hay bale, hay stack, (mud wrestling:1.5), (sweaty:1.5), (drenched in sweat:1.4), (dirty:1.5), (mud:1.5), (muddy:1.4), mud fight, plot,
BREAK
rarity, pony, feral, female, unicorn, mare, blue eyes, (dominant:1.3), on top, (lying on top of someone:1.2), looking down, evil grin,
BREAK
applejack, pony, feral, mare, female, earth pony, on bottom, prone, fighting, snarl, frown
I consider AutismMix and its derivatives to be overall better than Pony Diffusion V6 (used in PurpleSmart.ai), even if somewhat rigid.
Sure, you can use img2img with artist’s LoRA to get this effect, but I’m not sure how it would look, I haven’t experimented with artist LoRAs yet.
I never used autismmix(better than purplesmart?), but one of the things I wanted to do with AI is take a picture and apply an artists style to it.
BREAK
rarity, pony, feral, female, unicorn, mare, blue eyes, (dominant:1.3), on top, (lying on top of someone:1.2), looking down, evil grin,
BREAK
applejack, pony, feral, mare, female, earth pony, on bottom, prone, fighting, snarl, frown