AI art news and inspiration – 22 Dec '22
Featuring new Stable Diffusion techniques, 3D humans, better AI text, text-to-point cloud, 24/7 generated music streaming and interviews with Ganbrood and VJ_DMTFL.
Hey, thanks for being a subscriber to The Checkpoint over the last year, I really appreciate your support. Revue, the newsletter service I used before, has been Elon'd so it's taken me a while to get set up on a new platform. I hope you enjoy the new format.
This edition features new Stable Diffusion techniques, 3D humans, better AI text, text-to-point clouds, 24/7 generated music streaming and interviews with Ganbrood and VJ_DMTFL.
Have a great Christmas and new year!
News and updates
Stable Diffusion KLMC2 Animation – New notebook from Katherine Crowson which creates gentle transforming animations from a single prompt.
Riffusion – Fascinating use of Stable Diffusion which creates surprisingly listenable music from a model finetuned on audio spectograms.
Automatic1111 Stable Diffusion Photoshop plugin – Use Automatic1111 directly in Photoshop with this brand-new plugin. Not going to be long before SD and others are core features of any image editor.
Deforum Diffusion for Automatic1111 – I didn't know about this before, but it's possible to run Deforum as as script inside Automatic. Handy!
3DHumanGAN – Creates 3D views of full-body humans that are consistent across different angles and poses.
Point-E – New model from OpenAI that can create 3D point cloud models from text prompts.
Character-Aware models improve visual text rendering – New approach from Google Research that significantly improves the quality of AI-rendered text.
Bas Uterwijk discusses the challenge of using AI tools creatively, what makes a Ganbrood a Ganbrood and why AI art will be the defining art movement of this century.
VJ_DMTFL talks about the parallels between AI art and Vaporwave, developing their process and the feeling of awe when creating something new.