Action Poses

For those new to ControlNet and wondering how to use these files, the best tutorial I've found is this one (reddit).

Action poses for controlnet.

In the .zip files: .png to use in openpose and depth models. Added .json for openpose too so you can make modifications to the pose file. A more accurate description of what's in the .zip file is in each version description.

I am open to suggestions for what other categories to add here.

To do list:

• Aiming/shooting: done

• Fighting (punching/kicking): done

• Jumping: -

• ...

Additional notes:

• Use attributes like "from_behind", "from_below", "from_above", or "from_side" in your prompt to create the angles you are looking for.

• Climbing: several of the climbing poses work well with "from_above".

• Aiming/Shooting: SD doesn't generate hands nor guns well. I'd suggest using depth rather than openpose for aiming/shooting poses as you're at least getting an approximation of the shape of the weapon. You can also modify the depth .png in PS or paint to better approximate the shape of the character and weapon.

• Fighting (punching/kicking): Obviously requires inpainting. Photobashing also useful to add characters (targets) to your scene.

• Lowering the "Control Weight" and "Ending Control Step" parameters can help in creating poses that are slightly (or significantly) different from the origin pose. By my experience, for depth maps you have to go as low as 0.3 and 0.3 to see significant changes.

All pictures were generated with the Dreamshaper 7 or 8 Model.

Loras used: 3D Rendering Style (https://civitai.com/models/73756/3d-rendering-style) v1.4.1

Generation data: see attached pictures.