This is the workflow or this video tutorial:   https://youtu.be/DnPvBPKbeAQ  

In this workflow, we see how random noise can be used either as a ControlNet map, or as a source for Latent injection, helping us creating better, more interesting results when compared to baseline text to image generations.

Want to support me? You can buy me a coffee here:  https://ko-fi.com/risunobushi

This workflow can be used by:

- using a simple Perlin noise to act as a Depth map for a ControlNet (or any kind of controlNet that accepts black and white maps as inputs);

- using a mix of Perlin and gradient to better influece Depth;

- using a mix of Perlin, gradient, and freehand sketches to refine the shapes and areas we want to condition;

- using a mix of random noise and color pallette to create a noise-injected Latent;

- using everything together, both noise-injected Latent and a noise-driven ControlNet.