Spaces:

sd-concepts-library

/

stable-diffusion-conceptualizer

like 574

Running on A10G

I cannot get it working at all, can I use any image from the internet like from my own website?

I tried using a 100 images for the style conceptualization but the result is either an all black image or simply bad.
Although using the same style, the 100 images are pretty different.

I used 6 with good results.

I used 4 and I am unhappy with the results. The artist's style wasn't captured.

I used 4 and I am unhappy with the results. The artist's style wasn't captured.

Yeah this web version is crap, but when i run it locally it works like a charm

I used 4 and I am unhappy with the results. The artist's style wasn't captured.

Yeah this web version is crap, but when i run it locally it works like a charm

So you mean the colab version sucks?

https://huggingface.co/spaces/sd-concepts-library/stable-diffusion-conceptualizer

This one works pretty bad for me, but as i said running a local version on my pc with the downloaded styles works perfectly.

https://huggingface.co/spaces/sd-concepts-library/stable-diffusion-conceptualizer

This one works pretty bad for me, but as i said running a local version on my pc with the downloaded styles works perfectly.

How does that work locally?

Also can you share some prompts, input images for training and results?

Also can you share some prompts, input images for training and results?

I tried in colab training a style with 750 images, and it works well. However, it took about 4h of training.

Someone is saying that the output images are black. This happens when Stable Diffusion detects an image as NSFW. This filter can be disabled, but then is your responsability to not publish any image that do not follow the intended use of this model.

Can someone please explain how to run this locally? I am a noob . Thanks in advance

Can someone please explain how to run this locally? I am a noob . Thanks in advance

I have not tried it, but you can check the Stable Diffusion repo at Github. You can clone it on your local device, install all the necessary packages with $ pip install require ments.txt, and then run the code as explained in the readme.md file.

I have some technical question, I am not a programmer but an artist. I have images that I want to use on google drive, how do I link the drive folder to colab instead of image urls?

6 images 10000 steps is good.

I have some technical question, I am not a programmer but an artist. I have images that I want to use on google drive, how do I link the drive folder to colab instead of image urls?

From the Colab interface there is a button that says "Mount Drive". If you click on it, you will be able to access all your drive files from the Colab environment, just like if you had uploaded any other folder into the environment.

Can someone please explain how to run this locally? I am a noob . Thanks in advance

I have not tried it, but you can check the Stable Diffusion repo at Github. You can clone it on your local device, install all the necessary packages with $ pip install require ments.txt, and then run the code as explained in the readme.md file.

Wow, great way to describe something to a noob! I have no idea what my local device is, or a package pip or whatever! Please explain more thoroughly!