I didn't know who Harry Styles was until today when I learned that he's the guy who blocked off the entire St. Abbs harbour and prevented us to enjoy our fresh crab rolls. 😫🚧🕺🦀🥖
Another attempt at a longer piece. An imaginary Jerome K. Jerome writes about Twitter. All I seeded was the title, the author's name and the first "It", the rest is done by
#gpt3
Here is the full-length version as a PDF:
I started collecting some direction vectors in
#StyleGAN2
that try to change only a certain property of a generated face. But it's hard to get clean ones. Here's head rotation:
So spectral normalization is a mixed bag. Overall it seems to make everything more consistent and there are no blatant artifacts anymore. But it couldn't fully heal the one that had settled in the top-left corner.
I am proud that I had the chance to work with
#MassiveAttack
creating neural visuals for their new
#Mezzanine
anniversary tour which started yesterday. (I ripped that video from in_cmmn on instagram).
I do not really care much for "pretty" generative art. I want something that has an interesting concept, is algorithmically challenging and ideally so complex that I cannot reverse engineer its mechanism right away. Unfortunately I don't see a lot of that these days.
Now that I have my local version of
#BigGAN
running, exploring the wilderness and weirdness of its latent space will become a bit faster and comfortable.
If you just woke up to and wonder what happened to the hic et nunc website and its "discontinued" twitter account here's some info what is going on as far as I know:
I've created an experimental GAN architecture I call
#RecuResGAN
or "Recursive-Residual GAN" and I am pretty astonished that:
- it works at all
- how well it works across a pretty wide range of scales.
- it is just 15% the size of a comparable
#pix2pixHD
model
Neural Filters is a major breakthrough in AI-powered creativity and the beginning of a complete reimagination of filters and image manipulation inside
@Photoshop
🤩:
#AdobeMAX
Another round of fine-tuning, starting from a state where the artifacts had not shown yet. This model uses batch norm in the deep residual layers, instance norm in the surrounding ones and spectral norm in the discriminator.
This might show it a bit better. Left is input, right is output. My latest progress in neural enhancement using a modified
#pix2pix
architecture which tries to achieve better details. The training data right now is female faces only.
No, "Famous artist finding a new way to make more money in an already booming market" is not a game changer.
"Artists from all backgrounds creating a thriving decentralized ecosystem and community from scratch on the blockchain without having any funding" is one.
@hicetnunc2000
Josh Mobley
@joshmobleymusic
just sent me a new song called "A Walk in the Park" and I had my algorithm give it the
#BigGAN
treatment. Since it can create an infinite amount of versions it's hard to pick one, but I really like what it does at 1:08.
It's no secret that Art Basel Miami is all about NFTs this year. Make sure to check in at the
@tezos
booth and get your surreal self-portrait NFT minted by my installation:
@TheArtNewspaper
After a few hours of figuring out how to translate bezier curves into robot moves that don't wiggle, stutter or run into singularities I finally managed to have it sketch a
#CLIP
generated "Mona Lisa".
I had to try my own method for this problem. Not sure if you can call it an improvement, but by simply starting the gradient descent from different random locations in latent space you can already get more variation in the results.
Some
#genart
history:
Algorithm 133 by Peter G. Behrenz from 1962 might be the first documented random number generator used in generative art.
It can be found in Georg Nees' book "Generative Computergraphik"
WARNING: Unfortunately somebody has figured out a new nasty copyminting scam on
@hicetnunc2000
which is very deceptive since it shows the real artists as the creator, but they did not actually mint it.
Do not buy from this wallet: tz1ch5WHJDZbbFR8R98cDkriVMwFP6RN9it2