Here’s about 1/4 the line for this event 👀

The next wave of physical AI from NVIDIA Cosmos. A world foundation model trained on 20 million hours of video that understands the physical world. 🤯

Made a new friend

Video-to-Video in Runway is nuts. I tried giving it a challenge and it’s incredible it was able to track all the balls so well. This was on the first try. I uploaded the source video then used just a simple prompt. A man is juggling fireballs in a dimly lit room. And that’s

First time combining video to video with Act One lip syncing in Runway a while back. So many fun possibilities with this

Breakdancing in Veo 3. Still wonky, but seems to have the most coherent wonkiness of any model.

Midjourney video is it's own kind of awesome. Just clicked animate on a bunch of old generations, but so far it seems to be really good at animating images other models don't know what to do with.

Nice one Midjourney

Generated an image with Midjourney V1, animated in Midjourney, then upscaled with Astra by Topaz Labs... For some reason

Cheers, Midjourney

iykyk

The Cinders Flux Kontext + Veo 3 is a killer combo for consistent characters with dialogue. The original characters and aesthetics were from Midjourney, of course

Didn’t even take the time to spell check. It’s probably nothing

First test with Grok Imagine. Why did I click spicy?🤦‍♂️

This type of map animation used to take days of complex motion graphics work in After Effects. I made it in under an hour using Nano Banana and Hailuo Actually blown away this was possible