robgon-art/GANfolk

GANfolk are AI-generated renderings of fictional people. Each image in the collection was created by a pair of Generative Adversarial Networks (GANs) with names and backstories also created with AI. The GANs were trained using portraits from artists like Renoir, Turner, and Modigliani in addition to open-source, modern photos.

GANfolk: Using AI to create portraits of fictional people to sell as NFTs

How to use GPT-3, StyleGAN2, and VQGAN to synthesize diverse characters from open-source images

By Robert. A Gonsalves

You can see my article on Medium and check out the results here, opensea.io/collection/ganfolk

The source code and generated images are released under the CC BY-SA license.

If you use this project to create images, please give attribution like this: This image was created with GANfolk by Robert A. Gonsalves.

Google Colabs

• GANfolk - StyleGAN 2 and CLIP
• GANfolk - VQGAN and CLIP

Acknowledgements

• StyleGAN 2 by T. Karras, et al., Analyzing and Improving the Image Quality of StyleGAN
  
• StyleGAN 2 implementation by rosinality, https://github.com/rosinality/stylegan2-pytorch
  
• VQGAN by P. Esser, R. Rombach, and B. Ommer, Taming Transformers for High-Resolution Image Synthesis
  
• CLIP by A. Radford, et al., Learning Transferable Visual Models From Natural Language Supervision
  
• GPT-3 by Tom B. Brown, et al., Language Models are Few-Shot Learners
• The original VQGAN+Clip notebook was made by Katherine Crowson, with further modifications by Justin John