Home Artists Posts Import Register

Content

The extension is already installed with the webui I use (vlad diffusion), but if you use Automatic1111 you need to install it yourself.

https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main Then download the model from here and place it in your controlnet models folder which is in the extensions folder.

To use it, open the dropdown menu in your webui and enable it. For the input image you just need a black and white scribble.

Here's the ones I made for this post:

And here's the settings:

I would leave the control mode to "My prompt is more important" and the "ending control step" to 0.5, so the only setting you should really play with is the control weight.

Some more tips:

- For easy poses you can use a pretty low control weight and a very rough scribble, for more difficult ones you might want to increase the control weight and ideally draw a better scribble too.

- Try to generate some images that are close to the pose you want (even without controlnet) and use those as reference to draw better scribbles.

- If the results are not great no matter what settings and prompts you use, try to redraw the scribble.

Once I find a good base image I then edit it with my usual workflow (https://rentry.co/stuffyai), the only difference being that I enable controlnet in img2img too if needed (with the same scribble and a lower control weight).

I attached the scribbles I drew for these 2 images if you would like to try them, the file names include the resolution and the keywords you should use. They are very flexible and should work with any character, prompt and model you want.

Let me know if you would be interested in a pack with a bunch of different poses maybe.

Files

Comments

Briston

Thank you for the guide! Marvelous job!