Pinned Tweet
Check out this fun experiment! I was curious if I could build a dynamic semantic understanding of World Foundation Model based 3D scenes.
This pipeline captures the current view, runs object recognition, adds additional multi-modal enrichment, then converts 2D detections into o the 3D scene. I also made these "classification markers" searchable/navigable.
@theworldlabs @sparkjsdev @threejs and running CV on @awscloud .. Using AWS's Rekognition + @Google @GeminiApp for multi-modal CV and thinking to add @Meta's Segment Anything Model (SAM).
The idea was expanded from a demo I saw from @bilawalsidhu, which was based off a demo that @alexanderchen built.
What should I do with it? How should I expand it? This one was fun to build!
This pipeline captures the current view, runs object recognition, adds additional multi-modal enrichment, then converts 2D detections into o the 3D scene. I also made these "classification markers" searchable/navigable.
@theworldlabs @sparkjsdev @threejs and running CV on @awscloud .. Using AWS's Rekognition + @Google @GeminiApp for multi-modal CV and thinking to add @Meta's Segment Anything Model (SAM).
The idea was expanded from a demo I saw from @bilawalsidhu, which was based off a demo that @alexanderchen built.
What should I do with it? How should I expand it? This one was fun to build!
10
13
101
11,384
HAD TO DO IT! I love a good cloth simulation. This is the web!!! at 60fps.
I had to run some tests with WebGPU to see how far I could push it.
Give it a go below! And I forgot to mention in the video to go play with the cats.
@WebGPU @typeGPU @ElevenLabs @OpenAI Codex
l
I had to run some tests with WebGPU to see how far I could push it.
Give it a go below! And I forgot to mention in the video to go play with the cats.
@WebGPU @typeGPU @ElevenLabs @OpenAI Codex
l
0
0
1
14
I was Beta testing @SpAItial_AI's Echo-3.. Really incredible model and the mobile app makes capture very easy.
Miles ahead of other mobile capture apps that i've tested and I think I've tested almost all of them.
Miles ahead of other mobile capture apps that i've tested and I think I've tested almost all of them.
Introducing Echo-3. A state-of-the-art world model.
Echo-3 turns phone captures into photorealistic, metric-scale 3D worlds. From a single room to multi-floor buildings, outdoor and public spaces.
Available today.
1
4
21
2,911
I built the @Windows95 screen saver this weekend with @WebGPU for fun.
Now you can be inside it and run around. A little Monday morning nostalgia for you! Have fun!
Link in the comments
Now you can be inside it and run around. A little Monday morning nostalgia for you! Have fun!
Link in the comments
5
1
7
574
Awesome to see WebGPU by default. That escalated quickly 👀👀
🇨🇭 R3F v10 alpha 6 is out now! 🇨🇭
We took your feedback from @threejsconf and brewed a fresh alpha. Our biggest yet with 200+ commits thanks largely to @DennisSmolek. Turns out a major version is a good time for a few major changes.
✧ WebGPURenderer by default
✧ TSL hooks get -three/tsl
✧ cleaner imports, only bundle the renderer you use
✧ share one renderer across canvases with a single prop
✧ better React 19 support, smoother hot reloads and cleanup
✧ the scheduler gets fixed timestep physics
For a WebGLRenderer-first experience import your Canvas from react-three/fiber/legacy and keep going.
All your feedback is what keeps this moving forward. Give it a spin and tell us what breaks.
Release notes 👉 github.com/pmndrs/react-t…
Example 👉 pmndrs.github.io/examples/examp…
0
0
1
103
I wanted to explore typeGPU for webGPU, so made a little vehicle explorer and deployed it to adamwebsite.com/experiments/ty…
Really nice clean utilities, I like the foundation. I recommend a bookmark for later!
Check out these other examples: docs.swmansion.com/TypeGPU/exampl…
@iwoplaza
Really nice clean utilities, I like the foundation. I recommend a bookmark for later!
Check out these other examples: docs.swmansion.com/TypeGPU/exampl…
@iwoplaza
0
1
1
241
This in honestly amazing. Congrats @mrdoob and the whole community that contributed... Things are going to get interesting from here.
24M npm downloads / week 🚀
0
0
0
201
I got access to the @theworldlabs Atlas Beta!
Some fun initial experiments and thoughts here. And huge congrats to the World Labs team and @AMD on the acquisition. It'll be interesting to see where that goes!
Some fun initial experiments and thoughts here. And huge congrats to the World Labs team and @AMD on the acquisition. It'll be interesting to see where that goes!
1
0
19
5,993
For all my 3dgs nerds out there. Any idea on what this is? I thought corrupt file, but seems to have to do with distance from origin. Looks kinda cool
1
0
4
93
This is really cool. I love seeing experiments like this!!
This is getting ridiculous.
You need to try this AI filmmaking workflow.
My Blender reference is embarrassingly simple. You can drive a precise multi-character shot with nothing but basic shapes.
1. Generate a start frame in Midjourney
2. Block out the scene in Blender with ra
3. Feed both to Seedance
This feels like a real unlock. Your 3D reference doesn't need to look good.
0
0
0
128
360 Depth Map to World Foundation Model Pipeline!
New experiment! Low fidelity in, high fidelity out. These low poly to WFM workflows will be important for keeping granular control of 3DGS outputs.
I'm also thinking about extending this to automate the creation of tileable er. "World Tiles"?
Thanks to @XRarchitect for pointing me to the depth map API input.
@theworldlabs @awscloud @sparkjsdev
New experiment! Low fidelity in, high fidelity out. These low poly to WFM workflows will be important for keeping granular control of 3DGS outputs.
I'm also thinking about extending this to automate the creation of tileable er. "World Tiles"?
Thanks to @XRarchitect for pointing me to the depth map API input.
@theworldlabs @awscloud @sparkjsdev
1
1
17
1,494
Adam Chernick retweeted
Ported Google's Draco decoder to pure JavaScript.
mrdoob.github.io/draco.js/
4.3× smaller than the WASM build, byte-for-byte identical output, often faster once you factor in load, init and parse.
mrdoob.github.io/draco.js/
4.3× smaller than the WASM build, byte-for-byte identical output, often faster once you factor in load, init and parse.
52
125
1,545
95,627
Safe to say, 4DGS + World Foundation Models have a lot of potential.
This was a fun experiment combining streamed 4DGS via @gracia_vr's WebSDK and World Foundation Models via @theworldlabs into one cohesive experience. Fully navigable with re-lighting and spatial audio, in the e going to get weird.
Try it out here: adamwebsite.com/experiments/4d…
[WASD to fly and right/left arrow to jump between scenes]
Deployed on @awscloud with @sparkjsdev
This was a fun experiment combining streamed 4DGS via @gracia_vr's WebSDK and World Foundation Models via @theworldlabs into one cohesive experience. Fully navigable with re-lighting and spatial audio, in the e going to get weird.
Try it out here: adamwebsite.com/experiments/4d…
[WASD to fly and right/left arrow to jump between scenes]
Deployed on @awscloud with @sparkjsdev
1
1
8
645