Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Drag Your GAN: Interactive Point-Based Manipulation of Images

vcai.mpi-inf.mpg.de

Post reply on HN
Print view View original

21–30 of 90 posts

Re: Drag Your GAN: Interactive Point-Based Manipulation of Images

#23

Post by orbital-decay »

vagabund wrote: post #15

The semantic understanding feels much richer than diffusion based modeling, e.g. the trees on the shore growing to match the manipulated reflection, the sun changing shape as it's moved up on the horizon, the horse's leg following proper biomechanics as its position is changed. I haven't gotten such a cohesive world model when doing text-guided in-painting with stable diffusion etc. This feels like it could very conc…

Temporal consistency for a guided scene is a separate problem. It's been kind of solved a couple years ago. [0] It can be used with animation rigs and simplistic tagged geometries, and it even works in near real-time. "Kind of" because training a model from scratch from a large dataset is not something you want to do for the actual job; what you want is a good style transfer mechanism that can extract features from as few references as possible.

[0] https://isl-org.github.io/PhotorealismEnhancement/

Re: Drag Your GAN: Interactive Point-Based Manipulation of Images

#26

Post by foota »

bick_nyers wrote: post #10
imranq wrote: post #9

Earlier quoted context omitted.

There are probably ways to embed cryptographic hashes within images. Any device that creates images from the real world could have secret keys that can be used to validate any image created by said device. We will still need a centralized party that holds the secret keys for validation through

What about the recording device itself? Refeed a video/frame/hash back to a security camera and it tells you if it was originally sourced from that specific camera or not.

The sci-fi dystopian answer would be entangled photon lights and off site image recording that preserves entanglement :-)

Re: Drag Your GAN: Interactive Point-Based Manipulation of Images

#28

Post by ultra_nick »

tikkun wrote: post #22

Has anyone built the "online photoshop that incorporates all of the latest AI image editing tools asap and sells access as a premium subscription with lots of GPU access for smooth editing" business yet? I'd be curious to know.

Isn't that stability.ai's business model?

Re: Drag Your GAN: Interactive Point-Based Manipulation of Images

#29

Post by yamazakiwi »

u385639 wrote: post #27
loandbehold wrote: post #25

What do we need human actors for at this point? Everything can be generated by AI now.

Please try to make an original movie that meets the standard of, say, The Godfather, with AI.

I understand your point but I think it would be easier with AI than without. Many movies are not made to the standard of The Godfather because they don't sell like MCU Movie #53 and if you include more humans in the creation you're more likely to run into the current system's restrictions.

Making a movie as beloved as the Godfather would still be challenging of course.

Re: Drag Your GAN: Interactive Point-Based Manipulation of Images

#30

Post by apodolny »

tikkun wrote: post #22

Has anyone built the "online photoshop that incorporates all of the latest AI image editing tools asap and sells access as a premium subscription with lots of GPU access for smooth editing" business yet? I'd be curious to know.

Playground AI (https://playgroundai.com/) does a lot of this.
Post reply on HN