//=time() ?>
My video init for #stablediffusion animation. She doesnt look this good after 50 frames.
#AIart
I understand GAN models are trained in pixel image formats, but I look forward to a way of modifying and iterating vector art, running as a script in Illustrator or similar apps. Frustrating attempt with init image in Stable Diffusion.
An experiment with video_init from Little Otik by Jan Švankmajer. In this scene Otik's parents begin to be concerned about his appetites. Prompt: "An aesthetic ultra detailed film still, artwork by studio ghibli and pixar, makoto shinkai, pixv, 8k" #AIart #stablediffusion
Following @KyrickYoung steps, I asked #Stablediffusion to tell me what it sees in the Rorschach test using an Init image and an open prompt.
@prehis_wit_kris @arvalis @GroucuttJamie @Aleriia_V there's a point where you can no longer claim it's yours
only init img is mine in this case
The third step was use this image as init for Stable Diffusion. I did around 250 runs in total. Here some of the first results.
Daaaaaamn what excited and amazing results right? Love this one click masterpieces 🤯!
Tried using my own art of a sun god as an init image in #stablediffusion , trying different settings and samplers.
You can use stable diffusion inits like a prompt driven style transfer!
Here I take an init and apply sketch and watercolor styling with a light 0.4 strength!
Prompts in the ALT. Check the 🧵for more examples 👇
Init's that I thought would really suck for disco, turn out to be really cool in deform stable. #AIart
Ghost Rider
init by #midjourney
render by #discodiffusion
#aiart #atoll #contemporaryart #creativecodeart #digitalartist #digitaldrawing #digitalillustration #draw #hiking #illustrator
fallen angle
init by #midjourney
render by #discodiffusion
#aiart #atoll #contemporaryart #creativecodeart #digitalartist #digitaldrawing #digitalillustration #draw #hiking #illustrator
Stable Init images lead to amazing results! It feels like what I wished dall-e variations to do.
Prompt in the ALT. More in the thread 👇
I'm giving you one of my favorite prompts btw 😉
#stablediffusion
"itself, raytrace, a hyper - lifelike, photorealistic,
digital painting, aesthetic, artstation, intricate oil"
These were generated on my local RTX3090 at 1024x640, 90 steps using euler_ancestral sampler & for many of these I used an init_img to get the composition just right👌
ということでちょっときりたんのファインチューニングを実施。画像は公式イラスト。30 EPOCHぐらいでinit_wordはhuman、girl。結果を読みませてプロンプトいろいろ調整した結果を添付。近しいものも遠そうなものもあるけど、包丁要素=ヤンデレ の修正がかかっている気がする #stablediffusion https://t.co/6I0YnS7yGa