Archive (2017-2018)
Alongside the Progressive Growing of GANs Landscapes and Nude Portraits (2018), and the Balenciaga Fan Project (2018-2019)- this is the work I was making in 2017 and 2018.Art-DCGAN (2017):
The first time I saw AI generated images was in 2017 In 2017 I moved to work at nvidia and finally was able to get a graphics card to train neural networks with. I had seen GAN generated images the year before, in the paper "Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks".
I really wanted to try and generate landscapes, I wrote a wikiart scraper and published the scraper + some pre-trained networks on different painting genres on github here:
https://github.com/robbiebarrat/art-DCGAN
The repository was intended for people who didn't have access to a GPU to be able to experiment with GANs for artwork, with instructions for using one of the pre-trained models as a base, and fine-tuning the network on a smaller dataset of chosen images (seen in the landscapes -> abstract landscapes example, which required such short training you could do it on a CPU).
Neural Network as a Video Game Shader (2018):
Using CycleGAN, a corpus of minecraft gameplay frames I collected through my own gameplay, and a corpus of landscape oil paintings I scraped from WikiArt - I was able to train the network to turn minecraft into a sort of "playable oil painting". There's a twitter thread about this work here: https://x.com/videodrome/status/973268632233852928 - and another thread where I talk about the reverse process of going from oil painting to minecraft screenshot here: https://x.com/videodrome/status/995773501012299776
Generative Plant Sketches (2018):
This was a project focusing on using electrical signals from plants to work as input for processing sketches plotted on an axidraw (pen plotter). I was interested to see whether the electrical signals would differ enough during nighttime vs daytime (when the plant is getting more light) to influence the appearance of the sketch.
There's a tutorial and all the code + shopping list on my github repository for this project: https://github.com/robbiebarrat/plant-art/blob/master/README.md
Sculpture GAN / AI Generated Sculptures (2018):
3D DCGAN inspired GAN trained on 32x32x32 voxelizations of Thingi10k - a corpus of 10,000 3D printable objects and sculptures - as a result, the generated sculptures are almost always 3D-printable, but usually do not have any real 'meaning', and are just abstract sorts of shapes.
I originally began this project as an attempt at generative architecture; but lack of an appropriate dataset held me back from that.
The code is available on my github page for this project: https://github.com/robbiebarrat/Sculpture-GAN
I only ever printed one sculpture from this project, and in 2026 it found a home with my friend Archivist: https://archiv.ist/#sculpture-gan
Rapping Neural Network (2016/2017):
The first generative art project I did, using a recurrent neural network to generate Kanye West songs. I'm still incredibly gracious to Dave Gershgorn for writing a piece about this project on Quartz: https://qz.com/920091/a-west-virginia-teen-taught-himself-how-to-build-a-rapping-ai-using-kanye-west-lyrics - since this project becoming popular got me a job at nvidia immediately out of high school.
The code for this project is on my github here: https://github.com/robbiebarrat/rapping-neural-network
You can listen to an output of the program here: