Update: for those that want to run a local stable diffusion (text to image generator) setup, I highly recommend fastsd_cpu.
There are some requirements, but chances are you already have them installed.
- git installed
- pip installed
- qt5-qtbase-gui
Use git clone to get the repo here https://github.com/rupeshs/fastsdcpu
Open a terminal in the fastsdcpu folder you just cloned
Type pip install -r requirements.txt this part takes a little while
Once thats done go to fastsdcpu folder and right click install.sh, click permissions and make sure its executable. Then double click and a small window should pop up giving a choice to execute/execute in terminal, choose "in terminal", this part also take a little while
Once thats done you can double click start.sh, again it will give the same options as before again choose "in terminal"
Soon after a gui should pop up. If not and the terminal shows an error, it is most likely a qt issue ( figuring out the fix) in this case try using the start-webui.sh
If a browser doesnt auto open with the web ui you can open a browser and plug in the ip shown in terminal
On the first run its going to take awhile since it first has to download the chosen model.
I recommend the lcm-lora option its default lykon/dreamshaperv8 is pretty decent
Ive been using this for awhile and it works pretty well for what is and what it is running on.
For best results use the 1gb payload when booting linux, we want the most ram possible, also I would recommend an 8gb swap.
On 1gb payload a 512x512 image takes just under 20 seconds per step and I use a bare minimum of 4 steps per image
Here is a few Ive generated on ps4linux using fastsdcpu



If you have any questions, or any issues let me know and ill do my best to help.