𝐂𝐡𝐞 𝐆𝐮𝐞𝐯𝐚𝐫𝐚
chaynabors.com
𝐂𝐡𝐞 𝐆𝐮𝐞𝐯𝐚𝐫𝐚
@chaynabors.com
Finally, if it were me doing this work, I would want to stick with voxelized worlds for procedural content, or rasterized for statically authored. Very small chance I pick SDF for static author too depending on game. Don’t be like me and only write engine code and never sims! Forever prototyping…
October 21, 2025 at 2:11 PM
I assume you’re thinking to author soft bodies with sdfs, with arbitrary limbs or eyes. That’s very powerful and also enormously difficult. You can in fact approximate a bounding volume for these things depending on shader, you won’t need any other acceleration structures most likely.
October 21, 2025 at 2:04 PM
On that note, models using SDFs in this hypothetical would be like, a shader for a tree, with a seed in push constants if you can fit it, and every model becomes a coding challenge. Not impossible. Explicit data keeps the door open to prefabs, so voxels are still holy grail for generation imo
October 21, 2025 at 1:59 PM
I’m hesitant for many reasons. I’d suggest authoring just one SDF model you’re satisfied with on shadertoy first, and then scaling that up to solve the perf issues rather than working backwards from the open world perspective. Authoring is my biggest concern, but could work depending on subject.
October 21, 2025 at 1:55 PM
The acceleration structures fall apart if you want to start bouncing light around the scene in real time, at which point you’d be better off with an uber shader. You can’t frustum cull things you have to intersect against. You can offline calculate diffuse and SH coefficients if your scene is static
October 21, 2025 at 3:43 AM
A lot to unpack but bounding volumes for implicit volumes aren’t trivial to approximate. You could have a shader per object, and deferred rendering minimizes work with bounding volumes for lights similarly. Subchunking and light bounds allow for lighting across chunks of uniform size. I could go on…
October 21, 2025 at 3:36 AM
Yeah, also see github.com/rust-windowi....

CPU renderers are slow but could work with enough optimization, threading, etc.

If you can make WGPU work for you, you could speed up your automata in compute.

I’d only use WASM/WGPU for web. There are many ways to reconcile using different languages.
GitHub - rust-windowing/softbuffer: Easily write an image to a window
Easily write an image to a window. Contribute to rust-windowing/softbuffer development by creating an account on GitHub.
github.com
September 9, 2025 at 1:54 PM
Source: graphics is my passion 🌈
September 9, 2025 at 2:02 AM
What’s drawing you to WASM? You could use the pixels crate which has an underlying WebGPU backend, but’s it’s unlikely to be easy if you don’t know Rust. I suspect it would give you exactly the CPU feel and ergonomics you want, and they have a web example. github.com/parasyte/pix...
GitHub - parasyte/pixels: A tiny hardware-accelerated pixel frame buffer. 🦀
A tiny hardware-accelerated pixel frame buffer. 🦀. Contribute to parasyte/pixels development by creating an account on GitHub.
github.com
September 9, 2025 at 2:01 AM
Reposted by 𝐂𝐡𝐞 𝐆𝐮𝐞𝐯𝐚𝐫𝐚
Every day, thousands¹ of indie devs lament: "Oh no, I can't find a Carolingian minuscule that matches the resolution of my retro style game!"

Well, good news!
www.pentacom.jp/pentacom/bit...

¹ ᵉˢᵗᶦᵐᵃᵗᵉᵈ
May 22, 2025 at 8:41 PM
her name is celka and she loves her dad. she gets into trouble to get his attention because he’s so busy.
February 6, 2025 at 9:54 AM
I’ve been quietly following the development of these for years. Having used them, they’re really very good and you should absolutely charge for them. Sadly, my handwritten graphics pipeline graduated simple diffuse, but let me know if you’d like some scripts for bulk renaming, etc, all free.
January 27, 2025 at 9:37 AM