Andy Cheng

Builder @theworldlabs

[Andy Cheng @AndyCheng_JH]

Imagine what we can build when one model can index space and time to generate 3D-consistent worlds! After months of grinding with this incredible team @theworldlabs, it’s so exciting to see Atlas come together!

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[Kaifeng Zhang @kaiwynd]

Can point flow completion serve as a pre-training objective for learning to understand physics?

Excited to share PointZero, our recent work on bringing this idea to reality.

My favorite part is that the model works zero-shot on real-world images! Take a photo, and then start to simulate it with control points. An embodiment-agnostic, scalable approach.

Huge congrats to @bardienus and the team!

[Andy Cheng @AndyCheng_JH]

@thoma_gu 🤩

[World Labs @theworldlabs]

One more thing…

[Ziqi Ma @ziqi__ma]

Today’s video world models “simulate” the world by generating pixel frame observations🖼️. Can they continue to simulate the world when observations are interrupted - such as by occlusion, illumination dimming, or camera lookaway?
To probe this question, we release STEVO-Bench, which holistically evaluates whether image-/text-to-video models and camera-controlled video models can correctly evolve states under observation control. Check out our website, blog and paper for how they fail!

[Andy Cheng @AndyCheng_JH]

@ziqi__ma Great work!!

[Andy Cheng @AndyCheng_JH]

It feels amazing to see the bullet-time results come together, especially when we were all so close to giving up. Huge thanks to @jcjohnss and @BenMildenhall for trusting me!

[Hao Zhang @HaoZhang623]

Jump to 2:26 to spot me and the team @DrTunnels @davidpantera_ @ZeCh452318 behind the watermelon bullet-time shots 🍉🎥

[Fei-Fei Li @drfeifei]

Next view prediction is the key to Atlas, enabling us to unify pixel-level generation and reconstruction. @jcjohnss @BenMildenhall @martin_casado and I had a deeper discussion on some of the most exciting technical innovations of Atlas, our newly released world model for spatial intelligence!

[Keunhong Park @KeunhongP]

When we say Atlas has pixel-perfect camera control, we mean it.

Atlas precisely follows input camera parameters, including non-planar projections such as the Brown-Conrady distortion model and the Kannala-Brandt fish-eye model.

@BhamidipatiPan1 invented a novel method of camera conditioning and it works beautifully.

🧵 [1/N]

[Keunhong Park @KeunhongP]

@sheldonblewis interned with us this summer and made pivotal contributions to Atlas.

If you're looking for someone with superb execution and research sense you should reach out to him.

[Andy Cheng @AndyCheng_JH]

I’ve learned so much from working with @KeunhongP . He’s a true technical leader and a detail-oriented thinker who excels in technical leadership, implementation, and execution.

[martin_casado @martin_casado]

Keunhong has been a key contributor to the WL base models, including Atlas. You should follow him.

[Andy Cheng @AndyCheng_JH]

@kdexd @theworldlabs 💪🔥Amazing!

[Andy Cheng @AndyCheng_JH]

@bilawalsidhu I have recording of our rehearsal 🤣🤣

[Andy Cheng @AndyCheng_JH]

@bilawalsidhu 1 😎

[Andy Cheng @AndyCheng_JH]

@bilawalsidhu It was so much fun watching David do this live! 🤣🤣🤣

[Andy Cheng @AndyCheng_JH]

@bilawalsidhu @saji8k It’s not using gen lock. We even tried different version of iPhones (i13 i14,…) and motion cameras and they worked well.

[Andy Cheng @AndyCheng_JH]

@ahsu___ @theworldlabs 🙏🤩Thanks Adrian!!

[Andy Cheng @AndyCheng_JH]

Imagine what we can build when one model can index space and time to generate 3D-consistent worlds! After months of grinding with this incredible team @theworldlabs, it’s so exciting to see Atlas come together!

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[Andy Cheng @AndyCheng_JH]

@jcjohnss @3scorciav @kdexd @_mbanani 🤤😍

[Andy Cheng @AndyCheng_JH]

@davidpantera_ @drfeifei More than 10 min 🤣

[Keunhong Park @KeunhongP]

Atlas also supports native 3D outputs, beating all specialized models in sparse reconstruction.

Big shoutout to our super interns @HaoZhang623 @bardienus @DrTunnels!

[martin_casado @martin_casado]

🔥An incredible accomplishment. Think of it as a video model with full camera control. And the scene remains (nearly) 3D consistent. Built on a fully internal base model. There are many use cases, from video editing, to 3D reconstruction to robotics. Great technical blog too.