The BiRefNet-lite model really works insanely well on WebGPU - even with some 3800x5200 photo I'm seeing <4s processing time for selecting a subject in the photo
Bennett
@vantezzenweb developer creating open-source software and other projects
- GoPro joins AI bonanza with pivot into data centers as shares skyrocket 38% cnbc.com/2026/09/01/gop…
- Got inspired by @reczko_konrad's demo and turned the light injection into a Mac app that you can use as a virtual camera in online meetings. DepthAnything v2 runs incredibly smooth on my M4 Pro, ray-marched relighting is then applied with a Metal shader00:31Depth-aware light injection in TypeGPU I got a 448x448 monocular depth model down to ~8 ms on my M4 Pro across ~250 dispatches, which is fast enough to use in realtime :D Since the inference is written directly in TypeGPU, I can just feed the depth buffer straight into the
- 🍪 React Cookie Banner for @shadcn /ui: 🔨 No configuration: Wrap services in a "CookieService" component and the banner auto-manages them ⛴️ For anything that renders: scripts, YouTube embeds, etc. 👨🎨 Native shadcn/ui: your Button, Dialog, Checkbox and Switch
- AutoForm for @shadcn /ui now uses a custom registry so you install it with one command > npx shadcn-ui@latest add raw.githubusercontent.com/vantezzen/auto… Also, if you are developing components yourself, check out github.com/vantezzen/shad… to create custom registries!
