Took me the whole day, but the new HTML in Canvas is working in my #webgpu library.
The code has been working for hours, no errors, the missing piece was that I needed to reassign the texture after the copyElementImageToTexture call, because it didn't updated the reference automatically.
I can even select the text and copy & paste. That's the blue rectangle, the live selection!
#javascript #chrome #webdev #graphicsprogramming #computergraphics
Part of my recent progress on my library POINTS (https://github.com/Absulit/points)
These are two render passes. Background pass has an image. Foreground pass has a mesh and an image inside; mesh is scaled dynamically at runtime. The image preserves its ratio no matter the scale of the quad.
#webgpu #wgsl #graphicsprogramming #javascript #webdev #webdevelopment
#genart #creativecoding #digitalart #3d #cgi #computeshaders #shaders #generativeart
Another cool things is that I've learned a lot about js these last few years. I know there's a lot more to learn, but I still recall there first iteration of my library, it was super slow, but it worked! And I thought that maybe something like C# could increase speed. It was basically the same, except the #javascript was slightly faster. The real issue was that I never understood #webgl, but it got fixed with #webgpu which makes a lot more of sense to me than webgl
I saw a video yesterday about #threejs and how thanks to their agnostic approach to the renderer which is isolated, they were able to add new renderers like the #webgpu one. I was thinking how I don't do that in my library and also how that ship has sailed, if I wanted to do that for a different renderer in the future I wouldn't have any idea of what to do, which is kind of funny.
Meaning, my library is pretty stuck with webgpu, or as the kids say: coupled.
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.
I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
absulit
sɹɐʎA xɘlA 💻➡🗑
Thomas