Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Check out Ariticraft 🦾 - a highly efficient agentic system that generates articulated 3D assets fully automatically at scale! 🚀

235,670 görüntüleme • 5 ay önce •via X (Twitter)

23 Yorum

Elliott / Shangzhe Wu profil fotoğrafı
Elliott / Shangzhe Wu5 ay önce

With this pipeline, we created the Articraft-10K dataset with 10,000+ articulated 3D assets. Try it out and contribute your generated assets to our repo: We’ll actively maintain the dataset, and all assets are fully open-source.

Ethan vs Machines profil fotoğrafı
Ethan vs Machines5 ay önce

Robot intelligence = LOT of data. Articraft = LOT of auto-generated parts 🤖 I keep saying Canada's advanced manufacturing is done. No future. Jump ship.

David Bui profil fotoğrafı
David Bui5 ay önce

this is sick

Henry Ventura profil fotoğrafı
Henry Ventura5 ay önce

this is insane and giving me tons of fomo

Prasanna Venkatesan profil fotoğrafı
Prasanna Venkatesan5 ay önce

This looks like a genuinely different approach to 3D asset creation — curious how it handles complex articulation constraints (e.g. multi-joint characters vs. simpler objects)? Also, any plans to open-source or offer API access?

Elliott / Shangzhe Wu profil fotoğrafı
Elliott / Shangzhe Wu5 ay önce

We open-sourced everything (code, paper, assets, agent traces)! You can launch the viewer from the codebase and inspect all the assets:

Tarasha Khurana profil fotoğrafı
Tarasha Khurana5 ay önce

@Oxford_VGG so cool Elliott!

Neuro profil fotoğrafı
Neuro5 ay önce

@KushMetaX @homeMetaX @0xGmedia

Mentis 🇦🇺 profil fotoğrafı
Mentis 🇦🇺5 ay önce

WAT?! 🤯

AiDevCraft profil fotoğrafı
AiDevCraft5 ay önce

Coding agents writing simulation-ready joints is the part that flips this from a 3D paper to a robotics dataset story — the validation loop is the real product. Curious how often Articraft's executor rejects a generated joint vs. silently passing a physically-broken one.

Elliott / Shangzhe Wu profil fotoğrafı
Elliott / Shangzhe Wu5 ay önce

Good question! All agent traces have been recorded and released with the dataset, so it should be possible to check

AiDevCraft profil fotoğrafı
AiDevCraft5 ay önce

Releasing the agent traces flips validation from a vendor claim into an auditable empirical number — rejection rate becomes a real benchmark instead of a marketing one. Worth surfacing it as a leaderboard metric so future generators can be compared against the same physics ground truth.

Yo.Town 3D profil fotoğrafı
Yo.Town 3D3 ay önce

Hi Elliott, we used Claude to generate about 500 lines of Python CAD for an articulated violin, then edited it through pinned notes, exported STEP, and tried in Fusion 360.

Elliott / Shangzhe Wu profil fotoğrafı
Elliott / Shangzhe Wu3 ay önce

very cool! Code is one of the best representations when it comes to design and editing.

Saurabh Patil profil fotoğrafı
Saurabh Patil5 ay önce

This is amazing !!

Julian le Roux profil fotoğrafı
Julian le Roux5 ay önce

abs(wow)

Simone profil fotoğrafı
Simone5 ay önce

wow this's awesome!

Rhy profil fotoğrafı
Rhy5 ay önce

we're building the application layer for this - would love to connect!

KidFl4sh profil fotoğrafı
KidFl4sh5 ay önce

I know you won’t feel a single ounce of remorse when 3D modelers will lose their jobs.

DDaisy.Sunny | minicoin profil fotoğrafı
DDaisy.Sunny | minicoin5 ay önce

tjis agentic 3d at scale is actually crazy bro hehe

Applemint(stakeholder) profil fotoğrafı
Applemint(stakeholder)5 ay önce

😳😳😳😳😳😳😳

Red pill or blue pill? profil fotoğrafı
Red pill or blue pill?5 ay önce

is it funded?

Triangulated Eidolon profil fotoğrafı
Triangulated Eidolon5 ay önce

Brilliant dataset and pipeline! Do you plan to add texturising and PBR material generation to it? How is the scaling mechanism working now?

Benzer Videolar