Skip to content

Latest commit

 

History

History
93 lines (63 loc) · 3.47 KB

GUI_TUTORIAL.md

File metadata and controls

93 lines (63 loc) · 3.47 KB
Raw

Stable Diffusion GUI tutorial (for now Windows only)

Step 0. Download the packed GUI

To get the binary, download this (4 gb).

Step 1.

Extract the downloaded archive.

Also, to download the weights, go here, and download this:

Rename it to model.ckpt

Put the model.ckpt file to StableDiffusionGui\_internal\stable_diffusion\models\ldm\stable-diffusion-v1

run the 2) download weights if not exist.bat file to check if the weights are placed in the right location.

also install git if not installed https://git-scm.com/downloads

Step 2.

Your folder contents should look like this

Step 3.

run the

1) install.bat

file. It should produce the output as in the screenshot.

Step 4.1

PEACASSO_GUI) run peacasso GUI.bat It will launch a web interface on https://127.0.0.1:8080

See more details here


You will see the peacasso gui, which looks like this:

Step 4.2

Each file explained: SD_FAST) run vanilla txt2img.bat is used to run the txt2img gradio interface. after double-clicking it, go to step 5.

SD_FAST) run vanilla img2img.bat is used to run the img2img gradio interface. after double-clicking it, go to step 5.

SD_FAST) run vanilla img2img inpainting.bat is used to run the img2img inpainting gradio interface. After double-clicking it, go to step 5.
If you are limited by resources or want to generate bigger-resolution images, go to step 4.3.

Step 4.3

Obviously the same as 4.2 but slower, though allows to generate much higher-resolution images, see the comparison table.

Step 4.3

If you encounter this error, just press 'OK', it doesn't mean anything.

Step 5

After clicking one of the 6 .bat files, you should see output like this.
Choose the red link in 99% cases. Choose the blue one if you have troubles with the red one. They lead to the same page. Copy your link of choice and paste it into your web-browser. (Chrome Firefox whatever)

Step +. Gradio interface explained.

Once you open the page, you should see interface like this. It may vary because each mode has it's interface.

Though I think it's intuitive, I will explain some of the params:

  • ddm_steps: usally 50 is the fine value, lower it if you want faster results
  • n_iter: number of generated images, will appear in a grid.
  • scale: how much your prompt will influence the image, experiment with it
  • turbo: if you have memory errors, try disabling it.
  • sampler: different sampler may produce a bit different results.
  • speed_mp: the more the slower, but the bigger images you can produce. Typically 2-5 are fine values.

Other params are better to be left as-is.

If you have any questions, leave a comment here