Ganvatar

Adjust age, gender, and emotion of faces with AI

3 followers

Visit website
All of the portraits in this demo are generated by an AI model called “StyleGAN”. Using a technique we call "semantic shaping", we're able to change the age, gender, or emotion of a face.
Ganvatar gallery image
Ganvatar gallery image
Launch tags:
Artificial Intelligence
Launch Team

What do you think? …

Ryan Hoover
This is wild. This is like v2 of the viral site,This Person Does Not Exist.
Berkeley Malagon
Maker
Hi all! Excited to share this (trippy) demo with the ProductHunt community. All of the portraits in this demo are computer-generated by a machine learning model called “StyleGAN”. While most of the recent excitement around StyleGAN centers around its amazing ability to generate infinite variation (e.g. thispersondoesnotexist.com <3), the emergent semantics encoded in the latent space are impressive as well. For instance, faces in this space allow for some semantic vector math, reminiscent of word2vec’s “king - man + woman = queen” (https://p.migdal.pl/2017/01/06/k...). We can find the latent representations of, say, smiling people. We can then average them and create a new semantic vector that, when added to pictures of non-smiling faces, makes them all smile. Some possible applications: - Generation of assets for games - Customizing ad photography by region/demographics - Lifelike, custom avatars - Compression - Modeling longitudinal medical imagery - Zero-shot inpainting, super-resolution, etc Happy to answer any questions! Shameless plug: if you're interested in working on this stuff with us, contact us at hello@psl.com :)
Vincent Tang
This site is just serving up pregenerated images made ahead of time using StyleGAN?
Berkeley Malagon
Maker
@vincentntang good eye - that’s right. Since the random images and semantic vectors are fixed, it didn’t make sense to run an inference server (yet!)
Vincent Tang
@berkeleymalagon this is combinations without repetitions? n=10, r=3. so it's 120 images? Couldn't help but think of math when seeing this
Berkeley Malagon
Maker
@vincentntang nope, just one pre-generated pic for each spot on the sliders, so, 10*5*5 = 250 total images. of course, the number of graduations on the sliders was an arbitrary choice. but, watch this space ;)
Vincent Tang
@berkeleymalagon ah okay silly me, it's permutations with repetition. Didn't see the last two sliders were 5 options. I was looking at the images too, how'd you make the transitions so seamless here too? really nice design
Berkeley Malagon
Maker
@vincentntang there's a blurring effect that makes it blend nicely :)