Summary

The website provides a guide on using PoseMyArt with Stable Diffusion to create custom body poses for graphic novels or comics, utilizing ControlNets for precise control.

Abstract

The guide outlines a method for artists and creators to generate custom body poses for their characters using the PoseMyArt platform in conjunction with Stable Diffusion's AI capabilities. It details a workflow that includes selecting and adjusting a character's pose on PoseMyArt, capturing the pose, removing any reference lines, and then using the resulting image as a Control Map in the Stable Diffusion AUTOMATIC1111 WebUI to produce a final image based on a text prompt. The process is aimed at those looking for intricate control over character poses in their creative projects, emphasizing the importance of removing background elements to prevent interference with the output image.

Opinions

The author suggests that creators of graphic novels or comics will benefit from the ability to customize character poses.
The use of PoseMyArt's paid features for character selection is recommended for a more personalized experience.
The author encourages the use of remove.bg or similar tools to eliminate background interference from the pose image.
The guide promotes the use of specific ControlNet pre-processors like fake_scribble and openpose for optimal results, while advising against options like canny to avoid an exact replication of the control map's contour.
The author provides a subjective sample prompt and encourages experimentation with different text prompts and ControlNet algorithms to achieve desired outcomes.

CONTROLNETS, POSEMYART

How to create the desired custom body pose using Stable Diffusion and PoseMyArt

Quick Guide on creating flexible body pose using PoseMy.Art

Context

If you are creating some graphic novels or comics, you will definitely want more control in determining the pose of your characters. Here is a quick way to create your own custom pose. But before that, here is a sample output.

A sample

Sample created from a human pose mesh ( Generated using Stable Diffusion )

This is the prompt we used for the above image.

photo of a super hero , raising both arms , wearing a leather dress with logo on his chest , full body shot , intricate, masterpiece, highest quality

Workflow

Goto posemy.art and create your pose, (You can select your character (if you are a paid user , or use some of the available characters available in the Free Plan)

2. Click on the nodes / joints of the body to select the orientation, degree of rotation. Move the circles through the desired angles to rotate/ move the body part / limb.

Image created in posemy.art ( Screenshot, edited by Author)

3. Once you get your desired pose for the character, take a screen shot ( Use the button, or manually take one )

This image will have the grid lines / reference lines from the application. These will interfere with the output image. So you will want to remove them from any of the free online applications like remove.bg ( https://www.remove.bg ) .

Remove the background using any free tool.

4. Copy this images ( with the back ground removed ) to your Stable Diffusion app. Here we are using AUTOMATIC1111 WebUI .

5. Use this as the Control Map and also as the input image. And go ahead and create your output image based on the desired text prompt .

If you are new to ControlNets, we highly encourage you to see the tutorial on how to get started with ControlNets . Link here .

Here is a the prompt that we used for this example :

super woman flying in the sky with left hand raised up , view from above , birds eye view , looking from above, view of the back of the head

And you are all done ! Go ahead and experiment with your prompt text and the ControlNet algorithms.

Output generated using ControlNet and Stable Diffusion

Tips

💡 Use one of the following algorithms as your ControlNet pre-processor for best results.

fake_scribble
openpose

The other options ( particularly canny) tend to exactly replicate the contour of the control map, which we do not want. We just need the pose !

References

[1] posemy.art ( the app to create your skeleton pose )

[2] https://www.remove.bg — Remove the background from the images