sebastiankamph
patreon
RuinedFooocus indepth guide & detailed usage
🕑 Added 2023-12-20 12:22:21 +0000 UTC
Detailed guide to install and use RuinedFooocus for Patreon Subscribers. This is a text & image guide for the video guide here: https://youtu.be/7ncu4f-PFGI
https://github.com/runew0lf/RuinedFooocus
Scroll down on the page and find the download link to the latest version.
Download the package and uncompress it, you might need to install 7zip if you can’t uncompress .7z files.
After you’ve uncompressed the files and folders, start with run.bat file
RuinedFooocus will now auto download a base model, if you already have sd_xl_base_1.0_0.9vae.safetensors you can copy it to the \models\checkpoints\ folder to speed up the install.
OPTIONAL: Available for manual download here: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0_0.9vae.safetensors
After the download has finished, you might or might not encounter errors and be prompted to press any key to continue. Do so and then re-run the file run.bat
Your RuinedFooocus should now start properly and if your browser doesn’t automatically open a tab, you can copy/paste 127.0.0.1:7860 and input that in a new browser tab.
You should now see this window and your RuinedFooocus has started.
In the prompt window you can type your prompt, ie. what you want to generate. For example, a cat in a hat
Once you press generate, your image will be generated. You can drag and drop it from the little preview image to save it. It will also be saved in your \RuinedFooocus\outputs\ folder
For more features, press the checkbox in the lower left corner, which will expand all the advanced features of RuinedFooocus
The settings tab will have options for the image output size, style and number of images. The aspect ratio sizes available are presets that work well with the Stable Diffusion XL model.
You can select one or multiple styles to suit your image.
If you want to recreate an image with the same settings, uncheck the “Random Seed” box. You can then iterate on an image, by adjusting other settings.
In the models tab, you can select the model used for Stable Diffusion. I recommend downloading your own models from a model database like https://civitai.com
Make sure they’re SDXL models and place them in the \RuinedFooocus\models\checkpoints\ folder.
You can also select Loras here and any loras you download will have to placed in the \RuinedFooocus\models\loras\ folder. Adjust the weight (how much strength you want to give the lora) with the slider.
OBP is One Button Prompt and is a great source of inspiration, you can press “Instant OBP” for a random prompt, or use the dropdowns to create your own randomized prompt with preset variables of your choosing.
These longer prompts are generally not required for SDXL, but One Button Prompt is a fun way to achieve new results.
EVO is an experimental feature where you get 9 results and can then select which image you prefer to evolve further from. You can do this indefinitely and keep evolving images. Start with a random prompt of your choosing and let it evolve into madness, or fantastic images.
In the PowerUp tab you have options like ControlNet, Img2Img and Inpainting. These use the same models as in automatic1111 but have less options to choose from to simplify the experience.
There’s also a custom option where more options will be present to you.
Canny, Depth, Recolor & Sketch are ControlNet models. They will use your input image as a reference and control the likeness of the output image.
Img2Img will use your input image as a reference and use similar colours in your output.
Upscale will simply make your image larger
If you select Inpainting, you will be able to paint a mask on your image and it will try to input your prompt in that mask. In the example, we’re prompting woman face, and will get a more detailed face. As you have no option for denoising strength, the inpainting results may vary.
The Info tab will show your settings for the current generated image.
Have fun generating beautiful images with RuinedFooocus!
Troubleshooting install:
Error: "MetadataIncompleteBuffer" - Your model files are corrupted. Please download models again.
Linux install
git clone https://github.com/runew0lf/RuinedFooocus.git
cd RuinedFooocus
conda env create -f environment.yaml
conda activate fooocus
pip install -r requirements_versions.txt
Let RuinedFooocus automatically download the models using the launcher:
python launch.py
Or if you want to open a remote port, use
python launch.py --listen
Comments
Sebastian Kamph
Not currently I believe
Tanvir Hafiz
does it have faceswap?
Xsin
is ruinedfooocus differente from fooocus?
Sebastian Kamph
You can use DPM++ 2m sde exponential or heun exponential in RF.
Don B
thank you for the great reference! Does RuinedFooocus support the DPM++ 2M SDE Heun Exponential sampler?
