HN Explorer

Show HN: We built an agentic image editor that preserves the original structure (palette.cam)

19 points by sakofchit ↗ HN
Hi everyone,

I’ve been experimenting with app where you can edit images in your camera roll simply by tweaking your photo’s metadata (changing location/time) and our agent will contextually regenerate the photo in that place & time in one shot. There's no prompting involved.

One of the hardest problems we’ve seen with these ai image editing/creation tools is that they struggle with preserving the subjects of the original image (faces, genders, number of people, bodies, animals, etc), and I think we’ve gotten a step closer to making it feel more realistic.

The gallery has some examples that people have been regenerating. https://palette.cam/gallery

Here’s a demo: https://x.com/sakofchit/status/1900274636522193067

Feel free to dm me on Twitter: https://twitter.com/sakofchit if you’d like to try out the TestFlight in the meantime

Would love to know what y'all think!

13 comments

[ 3.5 ms ] story [ 45.8 ms ] thread
jchiu1234
this is a pretty dope app!
rafram
Forgot to switch accounts?
jchiu1234
No, lol, I was forced to make an account for this. Mostly just a joke
kyt
Interesting project. What makes this an agent? Just looks like an image transform that uses LLMs.
jchiu1234
Good question, we designed a system that looks across the image and chooses which parts of the image to be preserved and which to be changed. And, if a region is selected to change, how should it be changed.

This is all done autonomously and the decision-maker is an agent.

zellyn
Why does it make random changes, like moving limbs, changing hairstyles (hilariously leaving a black person with dreads, but removing them from the white person beside them :facepalm:) etc?
jchiu1234
The diffusion model(s) the agent leverages to edit certain parts of the image are notorious for exhibiting weird behaviors (of course, we will improve these as we progress).

I'll let you figure out why it does some weird things regarding your comment (data distribution).

zoklet-enjoyer
That's a really cool idea. Reminds me of that AI camera someone made where it generates an image prompt based on, I think, GPS and weather data
guelo
On every photo in the gallery people's hairdos were vastly improved. That seems more useful than changing locations. You should pivot to an app that suggests better hairstyles for people.
sakofchit
interesting take! maybe there's something here as well, but i also do think there's a lot of ai apps out there that offer hairstyle changes