Behind
The wrestler is still gasping for air after that hefty slam. He's got mere moments before his rival sneaks up from behind again.
Rendered as artworks in the vein of American Realism - or at least that’s how the checkpoint model perceives American Realism, given the training data it was fed.
These paintings are based on a previously unpublished impressionistic painting that I made with text prompts in Midjourney (included as image 9).
I used the Depth control net to accurately extract poses and position from the painting. I prefer Depth over OpenPose for figure poses, as OpenPose's preprocessor can struggle with complex poses such as the one seen here.
For the settings, I use a control weight of 0.7 to exert a strong influence on the composition without locking it in place. Starting at 0 and ending at 0.5 give the figures flexibility to move and change their expression.
Figures sometimes change their position and expressions, which I can then control with a prompt. This is due to denoising what ended at 0.5, allowing for multiple adjustments.
Poses can be "locked in" as soon as 0.4, so if you desire to create a composition with given prompts, then ending at 0.4 is sufficient.
I've had conversations with a number of advanced Stable Diffusion users about Control Nets, and it often astonishes me that they don't utilize them at all. From these discussions, it dawned on me that these settings might not be readily apparent. Therefore, I intend to create some explanatory videos which I hope will prove beneficial.
If you're a newbie, I suggest you kick off by rendering at low settings and run an X/Y/Z plot for the weight, start, and end values. This will give you a clear idea of how they shape the final outcome.
Tech: Text prompts with Stable Diffusion: 30 steps, DPM++ SDE Karras, CFG 5, 512x512, Virile Fantasy v1.0, Denoising 0.5. ADetailer with prompts face_yolov8n.pt. Control Net Depth depth_midas. Hires 1.5x (768x768), 30 steps, 4x_foolhardy_Remacri. Post: Topaz Gigapixel HQ 4x (3072x3072). Adobe Lightroom. Adobe Photoshop AI Generative Layers Inpainting (limb fix).
