Paul Zhou

10 years working on AI research and applications

[Paul Zhou @chow459]

@elevenlabs @suno_ai_ @_akhaliq

[Paul Zhou @chow459]

🔥✨ Consistent Videos from Agent Generated 3D Prior

Check out our short demo film, crafted with our work on agent-based pipeline for generating controllable and physically plausible videos. (Unmute for the narration!) 🌟 Project link: https://t.co/6PsmcpFM5s

#AIGC 🌟🔥

[Paul Zhou @chow459]

We used @elevenlabs for narration and @suno_ai_ for soundtrack to make it more fun !

[Paul Zhou @chow459]

@multimodalart

[Paul Zhou @chow459]

@elonmusk Congratulations on passing the package

[Paul Zhou @chow459]

#LumaDreamMachine

[Paul Zhou @chow459]

Here are a few test cases I ran with @LumaLabsAI 's new Dream Machine. The consistency, camera motions are amazing. 1. Prompt: spiderman, running on a rooftop and jumping off, city night. Shows large camera motion that was not possible with previous openly available models.

[Paul Zhou @chow459]

in a ancient castle, basement, phantom's mask spins to camera

[Paul Zhou @chow459]

#LumaDreamMachine

[Paul Zhou @chow459]

Testing Luma AI on complicated physical interactions. Test 1: a group of dogs run towards the camera, running into each other and scatter, and run into each other, and scatter, repeatedly.

[Paul Zhou @chow459]

Prompt: a man opens a car's door, enters the car, and sits in the driver's seat. Analysis: model (or maybe prompt enhancement) has learned to split the scenes, avoiding the complicated physical actions of door opening, entering the car.

[Paul Zhou @chow459]

Image to video, prompt: man is very annoyed

[Paul Zhou @chow459]

An image to video example starting with a cartoon style image. Prompt: camera overlooks a dark but grand theatre