The Checkpoint #11 – The week in AI art

White roses / Pytti

What's new

ContraCLIP – An interesting StyleGAN+Clip model able to move between two contrasting sentence pairs, for example "a picture of a person with open eyes." → "a picture of a person with closed eyes." 

Parti pooper – Google announce another text-to-image model called Parti, which from the examples looks like it can create impressive results. However, sadly, they've decided not to make the code, models or data public.

CLIPMesh-SMPLX – Four text-to-3D models covering character design, expressions and poses.

Training custom AI generative models – Comprehensive collection of resources on training AI models by @pharmapsychotic.

Janelle Shane @JanelleCShane

Stunning transcript proving that GPT-3 may be secretly a squirrel. GPT-3 wrote the text in green, completly unedited! https://t.co/DwUjiXOZuY

Featured notebook – Pulp Sci-Fi Diffusion

Robot cat in a cyberpunk city

New from @KaliYuga_ai, a Disco Diffusion model fine-tuned on pulp sci-fi images.

Get it:

• Google Colab

• Github

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.