Since you didn't mention what problem you have, I'm guessing.
1) Artifacts around the masking edges: As mentioned in the guide
2) "Pornstar" breasts: see 6. Embeddings
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.
Since you didn't mention what problem you have, I'm guessing.
1) Artifacts around the masking edges: As mentioned in the guide
2) "Pornstar" breasts: see 6. Embeddings
1)As you told me to paint little bit outside the lines i did but can't get better results , it is getting distorted around the masked area
2) how can you mix the tone of the skin with actual skin tone for example the area you inpaint is slightly different from the original skin tone
1) try painting more outside of the edges. Sometimes I mask the whole body (in certain cases) to minimize distortion.
2) there is no ideal way to do this. The best option is to use the Sampling method: DPM++ SDE Karras , like I mentioned in the guide. That sampler seems the best at "copying" skin tones/colors to the masked area.
For the others, as I mentioned in the guide, if you don't provide details, like screenshots of the problem, or at least copy/paste of the docker output, I can't help. I'm not going to ask everyone individually to "provide more info".
for extra info here is the docker output:
and my data/stablediffusion file, I have tried with just to urpm files but the other files come back, I tried with all the files and it still won't work.
i have gotten it to work by opening the cmd in the sdwuidocker file and running the docker compose command, and it works fine.
but if i open docker desktop and stop the container and run it again with no changes I get the error
and my data/stablediffusion file, I have tried with just to urpm files but the other files come back, I tried with all the files and it still won't work.
i have gotten it to work by opening the cmd in the sdwuidocker file and running the docker compose command, and it works fine.
but if i open docker desktop and stop the container and run it again with no changes I get the error
Last edited:
If you look at the docker logs, you'll see:
I'm guessing the download of sd-v1-5-inpainting.ckpt did not complete, or got corrupted.
I'm unsure if the docker container will download it again if you delete it.
If it doesn't, go to , and download sd-v1-5-inpainting.ckpt, then go to and download v1-inpainting-inference.yaml and rename it to sd-v1-5-inpainting.yaml
Is this version of SD is better for inpainting? In comparison with normal SD?
what can i do to make this more effective and realistic? im really stuck right now beacause i cant get satisfiying results please how can i make realistic nudes. im using SD 1.5 and urpm.
Hey there !
Completely new to this kind of sh*t.
So far, each step is a challenge but I managed to troubleshoot myself.
Keep getting issue with 2nd code during 3rd step :
In the first time, I searched for the error code on the net and found my Powershell version was outdated.
Now running with the last version of Powershell 7 and the error still appears
Running on Windows 11 Family
10.0.22621 Build 22621
8Go RAM
AMD Ryzen 9 4900H
Thanks for your time mate !
Completely new to this kind of sh*t.
So far, each step is a challenge but I managed to troubleshoot myself.
Keep getting issue with 2nd code during 3rd step :
In the first time, I searched for the error code on the net and found my Powershell version was outdated.
Now running with the last version of Powershell 7 and the error still appears
Running on Windows 11 Family
10.0.22621 Build 22621
8Go RAM
AMD Ryzen 9 4900H
Thanks for your time mate !
To be clear: DO NOT use the standard Stable Diffusion 1.5 Inpainting model.. it's useless, trash, garbage, etc. The Stable Diffusion 2.x models are even worse, since they removed a lot of NSFW training data.
You need to select the URPM inpainting model (at the top left), as mentioned in the guide. You can also use "Realistic Vision 1.3 Inpainting" model available on Civitai.com, that one is also good for nude inpainting.
Additionally, if you don't give the app enough masking room, it won't work.. it can only 're-generate' the masked area.. so you need to mask a big area to let the model regenerate what is behind the mask, and let it work its magic. If you only mask a very small area, like only a 'tight' bikini, it won't work that great.
There were issues with this image (due to the lighting or something, not really sure) that caused a 'color distortion / difference' below the masked area.. it's still there, but a bit less since it masked more of the body.. i think the best way to solve this is to either mask her total body, or use another Sampling method. Below a full screenshot, including parameters (very basic) and the source & dest..
Result:
In most cases you will never replicate the full breast structure of the source image, due to the problems I listed above. So it won't be 100% perfect.
Please check the requirements, you need at least 32 GB RAM and a Nvidia graphics card is recommended.
I'm sorry, but 8 GB RAM will not work. Even 32 GB RAM is limited for this software, esp. if you have other stuff open, like a lot of browser tabs.
The mentioned 4-6 VRAM is graphics card RAM
I've added the requirements section also under 'non-GPU' and clarified that you probably need 64 GB of RAM if you don't have a Nvidia graphics card. But since I haven't tested it without a GPU, I've no way of knowing.
I have a RTX2060 with 4Go VRAM.
And since i already used softwares like Dreamtime to generate images with AI without any problem I did not find your statements concerning material requirements relevant in regard to my issue.
I will continue to look for a troubleshoot and will keep you updated with the result if I find a solution myself to contribute to the bug section ;)
Thanks for your time
I'm interested in running this but am not the most familiar with docker and powershell commands in general. I followed all the steps without issue through the setup until running "docker compose --profile auto up --build". This does finish the command and appears to build out the container for the webui, but right as it says the container webui is created and I presume the host should follow, I receive the following two errors:
Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: initialization error: load library failed: libnvidia-ml.so.1: cannot open shared object file: no such file or directory: unknown
I've looked through a few stackoverflow and docker forums to try and find an answer, but I'm not finding anything with similar inputs and the ones that might be applicable I'm likely not understanding how to translate to my situation. Would anyone be able to help or point me in the right direction?
Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: initialization error: load library failed: libnvidia-ml.so.1: cannot open shared object file: no such file or directory: unknown
I've looked through a few stackoverflow and docker forums to try and find an answer, but I'm not finding anything with similar inputs and the ones that might be applicable I'm likely not understanding how to translate to my situation. Would anyone be able to help or point me in the right direction?
saprtan ah, ok, your initial reply didn't include the gpu part, that's why I thought you had no gpu. I've searched for error code 137 and docker, and it returned the following:
So, unfortunately, I don't think AUTOMATIC1111's Webui will work with 8 GB of system RAM.
SeeRed do you have a GPU, and did you check the requirements ? You need recent Windows service packs:
More details:
So, unfortunately, I don't think AUTOMATIC1111's Webui will work with 8 GB of system RAM.
SeeRed do you have a GPU, and did you check the requirements ? You need recent Windows service packs:
More details:
I just tried to give it a go with some more RAM (16Go instead of 8Go originally) and #137 error still appears.
Too bad... Now it's time to buy new RAMs
Thanks for your time mate
During the download it said an error occurred. All path/URIs were downloaded other than GFPGAN. I received errors (errorcode 24) during the download about authorization failing. Is there a step of granting access that I missed?
Don't wanna be a dick or anything but have you tried just installing `python 3.10`
and just cloning/downloading the repo and clicking `webui-user.bat` on windows?
No windows pro required, no hyper-v? It should be lighter for your operating system
edit: why do you need docker and virtual machine anyway
Last edited: