Guide Installing Stable Diffusion Webui & Nudifying (Inpainting) Guide
- Thread starter softarchitect
- Start date
-
Welcome to the Fakes / AI / Deepfakes category
When posting AI generated content please include prompts used where possible.
While I have not tested it, this is my basis:
So as long as xformers is enabled (which it should be), 6GB of VRAM should be enough, at least when you leave resolution and batch size at the default 512x512 and 1.
can't get it to work. i'm pretty sure i've set everything up right and when i set the prompts, mask the image and hit Generate it immediately fails and gives this error:
Im not 100% sure if its my GPU which is really outdated (GTX 770) can someone confirm that for me
ive been googling a shit ton and can't find anything to fix this
Im not 100% sure if its my GPU which is really outdated (GTX 770) can someone confirm that for me
ive been googling a shit ton and can't find anything to fix this
Oh, I thought it was installed by default.
What you can try is to go to Extensions, Install from URL tab, and then enter and install.
Afterwards hit "Apply and Restart UI" from the Installed tab. That should normally install xformers in the docker container.
It also installs additional training functionality under the "Koyha sd-scripts" tab.
Unfortunately, it seems the GTX 770 has only 2 GB of VRAM ? The guide mentions that you would need a bare minimum of 4 GB of VRAM and even 4 GB is stretching it.
If your GPU does have 4 or more GB of graphics VRAM, you could try installing xformers as mentioned in my previous post.
But not 100% sure it will fix it in this case.
And you followed the fix as mentioned in the "Bugs" section of the guide ?
Last edited:
I’m new to this lol, can I just download the embeddings if I already have stable diffusion up and working. Or do I need to follow the whole process??
No, absolutely not. This is 'another' way to install AUTOMATIC1111's Webui. If you already have the Webui installed, you can just use the embeddings and/or models with your own install.
The folders probably are different, but it should work if you know where to place them.
If you use the docker version (as in this guide), you can put the LoRa models in <project folder>/data/Lora (should already exist). Then you go to the webui and click on the 3rd icon at the right (below the Generate button).
This will show all the embeddings/Lora available below the model parameter settings. You can click on "Lora" and should see all the LoRa models available, and if not, you can use the refresh button to reload them.
Yes
Source:
Generated Examples:
This is done with the Realistic Vision V13 inpainting model ( )
Last edited:
Yes, you just have to skip the docker / installation parts, but the rest should work. Just start from "The Webui"
Last edited:
6 GB VRAM "should" be enough, if not, you can install xformers and use the --xformers parameter which should bring the VRAM usage down. You could also use the --medvram or --lowvram parameters when starting the webui. More info:
But using medvram or lowvram move a lot to RAM, so you would need more RAM.
16 GB RAM is quite low.. you can try, but the minimum recommendation is 32 GB RAM, esp. when using medvram (docker default) or lowvram parameters.
But try it out, and let us know if it works, and the specific parameters.