I want to make a renderer no image based textures. To texture a model it's tessellated and each polygon is given color data. For flat planes, there would be a way to store those parts of the model data where each vert only has XY coordinates instead of XYZ

Email
Subject
Comment
Verification	Please type the name of the object/variant:
File
Embed
Voice
Poll
Password	(For file deletion.)

File: ClipboardImage.png 📥︎ (64.67 KB, 454x520) ImgOps

Chud 02/23/26 (Mon) 05:44:28 №27378 [Quote]

I want to make a renderer no image based textures. To texture a model it's tessellated and each polygon is given color data. For flat planes, there would be a way to store those parts of the model data where each vert only has XY coordinates instead of XYZ. Colors are indexed, possibly with different pallets for different subdivisions of each model. I'm sure other methods for both image and 3D model compression that already exist could be used for this.
Models are already very high poly and painting on a texture during the design process only to turn it into a flat skin in a massive niggerbillion sized atlas and map it back on in the pipeline has always seemed retarded to me.
This also opens up the possibility of vertex shading instead of per pixel, and giving each vertex a specular and/or alpha channel.

Chud 02/23/26 (Mon) 06:02:52 №27379 [Quote]

>>27378 (OP)
Now that I'm thinking, I'm wondering how color/alpha/specular is stored per tri. Vertices store position obviously but if they are ever shared then the tri count rises above the vert count. Maybe there is some elegant way around this, or maybe that limitation would still look cool.

Chud 02/23/26 (Mon) 06:03:29 №27380 [Quote]

one thrembillion tris per toothbrush award

Chud 02/23/26 (Mon) 06:26:22 №27382 [Quote]

>>27378 (OP)
I'm also very interested in pathtracing and how and applying how interlaced PNGs work in 3D but tris instead of voxels. I wish rendering could be clamped at one vertex per pixel or even have some density limit for 1 vert for n number of pixels on screen (this would shimmer a lot doe) and maybe foveation on top of that. Every time I try to visualize a solution in my brain I run into depth sorting or model stitching issues. I'm not proficient with any kind of topology or graphics math nor am I >150 IQ so idk if it's possible.