Graphics Noob

got soundfonts running in real-time in the browser! this gives the ability to produce some better quality/realistic piano sounds without generating a bunch of files will write some notes in this 🧵 before 😴, meantime try here! zz85.github.io/midiviz/017_sy… #wasm #rust #webaudio #js

@thdxr not otel, but i've got a ebpf cpu profiler wired to a tui flamegraph github.com/zz85/profile-b…

why has no one vibed up really good local otel tooling i want to be able to look at all these traces during local dev

@mrdoob funny and sad how many people fell for it.. at the same time that was a good way to see filter who is using AI to make posts

Every time I make an April Fools' I understand why Trump is president. x.com/i/trending/203…

@threejs love the highly optimized, hand-written low-level instructions to deliver 8x FPS performance! #L139C5-L139C76" target="_blank" rel="nofollow noopener">github.com/mrdoob/three.w…

10KB. 480+ fps. Faster than native. Try it: mrdoob.github.io/three.wasm Source: github.com/mrdoob/three.w…

The future of Three.js is WebAssembly

@mrdoob @stephomi cliff hanger! video cut off before i can see if you are sculpting a 🎳ball or a 😱 skull :)

Ported @stephomi's SculptGL to Three.js

in the screen capture, you can see toggling between a piano soundfont and the chiptune synthesizer, both pretty fun to listen to! if you're interested to know more, this youtube link would explain the music in more depth youtu.be/rQ81jzlR1y4?si…

Recently I've learnt that Alkan composed this piano piece (Etude in A, without Opus) in the 19th century that resembles 8-bit music or Rush-E ahead of it's time so I added this midi and a 8-bit synth (square waves) to my midi piano visualizer (🔊) Link: #C.V.Alkan%20-%20Etude%20sans%20opus.mid%7Cchiptune" target="_blank" rel="nofollow noopener">zz85.github.io/midiviz/midivi…

for retro mac versions, this site seems rather impressive infinitemac.org

went into a rabbit hole of booting up old OSes that run in webassembly in the browser. this one runs windows 2000 - copy.sh/v86/?profile=w… (pass: 1234) - i used another wasm iso creator to mount files on the vm ezyzip.com/create-iso-fil…

@mrdoob I took some screenshots... github.com/zz85/silly-ai/… the initial response also took more than a minute, most likely because it was crawling thru all your tweets

@BlurSpline I wonder what Grok responded (apart from Autechre and James Holden).

Thought this was the coolest video for a three.js release I've seen. so I'll generated some song/music celebrating r183 release (turn up🎧🔊 for best experience)

@mrdoob not sure how much AI really know your musical taste, but glad it's good entertainment :)

@BlurSpline 😂😂😂

@grok @mrdoob slightly more details/longer form write up (and the lyrics) posted to github.com/zz85/silly-ai/…

@grok @mrdoob 4. use ffmpeg to merge the audio with the release video. video has to be slowed 8x to sync with the music 5. used RIFE (Real-Time Intermediate Flow Estimation, a Deep learning model) for smooth interpolation 6. added motion blur using ffmpeg (for fpv-drone styled videos)

open source coding for me these days… create a bunch of issues/feature requests on my phone, assign issues to Claude (agents run in parallel in Github) . ends up with a bunch of PRs i’ve to review and verify, but coding using thumbs have a different meaning now…

@mrdoob that sounded like a genuine qn tho - that commenter probably doesn’t know the work involved using AI… nice you showed him the commit logs!

The Quake subreddit ended up approving my post and all the AI haters started downvoting 😁 reddit.com/r/quake/commen… A couple of guys reported bugs though 🙏

@xenovacom yay superb work! ~20 toks/s (medium) - ~30 toks/s (low reasoning) on Mac M4 Pro (48GB). x.com/BlurSpline/sta…

I get over 40 tokens/second on my M4 Max... what about you? Try it out yourself and let me know! 👇 huggingface.co/spaces/webml-c…

Behold... GPT-OSS (20B) running 100% locally in your browser on WebGPU. This shouldn't be possible — but with Transformers.js v4 and ONNX Runtime Web, it is! A new class of AI apps is emerging. Zero-install, infinite distribution. Simply visit a website and run models locally.

Not sure which quantized version, but I got about 22 tokens/s (medium reasoning) to 30 tokens/s (low reasoning) on a M4 Pro. For contrast, running GPT OSS 20B MXFP4 LmStudio (Mlx engine) gets me 60 tokens/s (low reasoning) but super impressive running on browser only + WebGPU!

I released Space-Radar v6 (my diskspace viz app) after 9 years after the last release. github.com/zz85/space-rad… New - apple silicon builds - memory viz for windows - electron bump - lots of other fixes (thanks @GitHubCopilot for making it easy to spin up agents for fixing issues)