The nexg big thing
Hi together,
maybe some of you already learned about AI generated art, but I want to let you all know, that this is a real game changer.
You can use Stable Diffusion to generate from primitive children drawings beautiful professional assets within seconds. Okay, should have at least a video card with 8gb ram.
Here is one example for a sword.
Your imagination just limits the possibilities.
Stable Diffusion Automatic1111 is one keyword for your search and you can find on Reddit with stable diffusion a lot of news.
And now you can give this site a new life. But you will still need some development, cause there is a lot of automatization happening at the moment.
there is also a model for 16 bit -pixel Landscapes art out there.
Kind regards
John2k
The original picture, beore I used the Image2Image feature.
Thanks for the information. We're aware of stable diffusion. There is already some AI generated art on the site. Would you be willing to give more detailed steps on generating Stable diffusion art? In it's current form, this tutorial doesn't have a lot of information on how people might get started with it
--Medicine StormHi Medicine Storm,
I miss the time to write a tutorial. And although Stable Diffusion got good GUIs, it's still a tool with a lot of parameters you can adjust. I can give some hints where to find all necessary information and already written tutorials.
You can run it online in several services. One of them is Google Colab. But I prefer the offline solution.
Personally, I started with the NKMD GUI. It's a one click solution. Very easy to use, but has not all features, like other solutions. https://nmkd.itch.io/t2i-gui
You can find the full feature list on the website. Runs only on NVIDIA video cards.
Automatic1111 is here actually the offline GUI with the most features. You can find it here:
https://github.com/AUTOMATIC1111/stable-diffusion-webui
In their own wiki you can find all information about the Gui and setup procedure.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki
When you want to be up-to-date, then Reddit help you much here:
https://www.reddit.com/r/StableDiffusion/
You can find there a lot of examples how people got the result, they posted and also tutorials. You may be need to dig a bit deeper, cause a lot of stuff happening there also with extended features, like training to achieve specific things.
I suggest you just to start with some ideas you have in your mind. With CFG Scale you can force the AI to follow more or less your input. Just experiment here. Standard settings usually produce good results.
Maybe you don't get good results, or it isn't close to something you imagined, then imagine how would you describe this picture to someone, who is blind? You can work with weights, do some prompt engineering.
Try Image2Image out. This is a really strong feature. Just paint something and put it in there. Don't forget to set the denoising correct, because maximum will mean, everything gets new generated, and something lower will follow more your image. Possibly you need some loop backs with new generated images, to refine the result. Of course, you can just send generated images also to image2image.
You can use outpainting to extend the image. You can use inpainting to modify or add something within your composition. Don't forget to switch for inpainting to the inpainting model and for image2image and text2image back again to the normal model!
Stable Diffusion was trained with a huge amount of 512x512 (cutouts?) of free accessible images, just to explain where sometimes strange compositions come from or extra arms, hands and so on. Don't wonder about much more limbs, fingers and heads. Also miss this model a lot of for us understandable things and sometimes the result will not be good, because it is uncommon. Then try Image2Image to realize your Idea.
This site is also nice, to have an overview over Stable Diffusion news:
https://rentry.org/sdupdates#newsfeed
Some videos on Youtube to get the Idea on how to setup/use it:
https://www.youtube.com/watch?v=mUpQMx9Qimg
From minute 21 he starts to explain some settings within Stable Diffusion.
https://www.youtube.com/watch?v=obVpFRTgQ8A
https://www.youtube.com/watch?v=c5dHIz0RyMU
Maybe also nice to know what for and where you can find available models:
https://rentry.org/sdmodels
I hope this rough summary can help someone to dive into AI image generation, and maybe someone else can write here a better tutorial.
Kind regards
John2k