Hello!
Akuma here and today with a realistic image. This image has been created locally on my computer with a graphics card that you never should use for AI stuff, because it is too slow, too old, too weak. BUT with some knowledge about Stable Diffusion and some patience and comprimises, you can still make it work. It is totally NOT! fun to work this way. You are extremely limited to the image size and steps the AI may take to create an image.
While most users just bang the "detail slider" up to maximum for every single image they produce, they don't understand how that actually works.
When I work locally with stable diffusion I can only work with 20-40 steps at max, depending on how detailed I do my prompts and what type of content I create.
So I challenged myself to create an image I can send to people and ask them how they like my photo (because people know I have a camera and love doing shoots). So far everyone that knew me didn't even think a split second I might just trick them and instantly liked the photo, one guy I know working for a model agency even asked me if she and her parents would be interested into a shooting for a german tv station.
This really made me happy, because I really love how this adorable smiling girl came out.
It took me about 4 hours of prompting and a giant prayer to our god Cthulhu that my VRAM won't be overloaded when face restoration kicks in. After about 40 - 50 images, I got this adorable smile.
Sadly I can not afford a GPU Server or a local 4090gtx to go extreme, but this is a good example that you still can do awesome images with extremely limited ressources.
Though I am still looking for a sponsor for a premium membership of an AI I would like to use, as I can not afford it, so if you really like my work and would be interested in supporting me, please DM on Discord "Akuma no Tsubasa#5151" (The System I want to use costs $30 a month).
If you would like to support me on a lower level:
https://ko-fi.com/akumasartificial
I am really EXTREMELY greateful for every tiny bit of support I can get as I sadly have to live of welfare and I am unable to work.
Thanks for your read and I really hope you liked the small journey through my little local experience of AI.
Hello!
Akuma here and today with a realistic image. This image has been created locally on my computer with a graphics card that you never should use for AI stuff, because it is too slow, too old, too weak. BUT with some knowledge about Stable Diffusion and some patience and comprimises, you can still make it work. It is totally NOT! fun to work this way. You are extremely limited to the image size and steps the AI may take to create an image.
While most users just bang the "detail slider" up to maximum for every single image they produce, they don't understand how that ac...