Made my own Perlin Noise and Perlin Power Fractal generator

Started by WAS, May 26, 2023, 05:40:34 PM

Previous topic - Next topic


I needed a (good) fast perlin generator for my WAS Node Suite for ComfyUI so I could make procedural textures to diffuse and stuff, so I spent some time learning stuff well above my head to create my own python perlin functions.

They don't actually use python for the heavy computation though, that's done with JIT and C.

Now if I can just do this for worley noise and not rely on a module for that. The original module I was using for perlin mode, was weird. Gave lots of calculation warnings like dividing by 0, and also produced a lot of dead generations (completely black). So this was much needed.

I actually eventually want add features to map generated textures to a sphere and use a path tracer engine that I've been looking at to render terrains and planets right within a stable diffusion pipeline. So far just building up all the noise functions and other stuff like warp noises, etc. Also need to figure out how to make these 3d/4d so I can traverse them. I theorize I can do that now, but still learning.


Cool, and the UI looks good. But why are you doing this, does it have something extra that's missing in other software?


It's stand alone diffusion software that's highly flexible. So can implement the latest technology very easily cause can just write the python nodes for it. I've written over 161 nodes for ComfyUI with my node suite:

Making it more familiar so I can do post processing and such right in it. Or in the case here, create displacement maps to diffuse out color textures for. The seamless displacement map below was made with my noise and blending nodes.

I also love that the entire workflows are in JSON format and stored in the image. So you can take that displacement image below, and load it in ComfyUI to get back the same workflow I used to create it, including the seed(s). It'd be cool if Terragen switched to JSON format and could use PNG format and PNGInfo or TIFF with TiffTags. Just load up a image with a node network right into TG. Haha