The Checkpoint - Issue #7

News & articles

Latent Princess Generator is now Majesty Diffusion – Lots of new features and a name change for this excellent notebook.

AvatarCLIP: Zero-Shot Text-Driven Generation and Animation of 3D Avatars – Text-to-image framework for creating and animating a 3D avatar. Github, Colab, Paper.

CAST – Domain Enhanced Arbitrary Image Style Transfer via Contrastive Learning – A novel approach to image style transfer that learns directly from image features, leading to a more consistent and less-distorted result. Paper.

Zoom, enhance! Using AI to enlarge images without losing quality – Super-resolution AI models are a new way of enlarging images without sacrificing image quality. Here's a comparison of the leading options as well as a brief background on how they work.

The Loris Emporium @EmporiumLoris

So, there has been a lot of "A.I. isn't Art" talk going around.. that you just push a button and BOOM, masterpiece. So I figured I would share a look into one of my workflows and explain a bit about myself and break down the creation process of "Leviathan" This be a 🧵 https://t.co/NbKEG9KRew

Featured notebook – Time-Travel Rephotography

Unlike other image restoration methods, though a clever use of StyleGAN, this model simulates what historical subjects would have looked like if photographed with modern camera equipment.

Instead of restoring the old image, it's projected into the latent space of contemporary photos, finding the closest equivalent and matching the pose and features.

The results are fascinating as they mix the recognisably old images with a modern portrait style, bringing the subjects to life in a way that a straight restoration and colourisation might not.

Get it:

• Website

• Github

• Colab

• Paper

Thanks for reading!

If you have anything you’d like to be featured or want to get in touch, give me a shout on Twitter or via email. Please also consider supporting me on Patreon so that I can spend more time creating content like this.