This week in multimodal ai art (23/May - 30/May)

Follow on Twitter, come hang on Discord, consider supporting on Patreon. Try MindsEye, our image generation GUI

* code released

Text-to-Image synthesizers:

- Google Imagen announced (Website, Paper)

by Google

Right after we released our weekly news last week Google announced Imagen - their state of the art text-to-image model that rivals Dall-E2. Check this twitter thread out here researchers generated images based on inputs given by people on Twitter

What do you want to imagen?

Please put the text below today and we'll try to get back to you with an image, subject to our internal review. #imagen https://t.co/7KRi3BGDXy
— Mohammad Norouzi (@mo_norouzi) May 25, 2022

- Improved Vector Quantized Diffusion Model released (GitHub)

by Microsoft

Microsoft released very stealthily (sent a new commit to the VQ-Diffusion repository) a follow up for this work. The paper is mentioned but wasn't released yet

Kind of stealthily Microsoft released "Improved VQ-Diffusion" - a follow-up on their technique that combines a VQ-VAE with diffusion

They released the code, the weights and a new VQ-VAE trained https://t.co/0GExuUJxOi

I'm running the first experiments: pic.twitter.com/QXS2lIvVlg
— multimodal ai art (@multimodalart) May 30, 2022

- Majesty Diffusion v1.3 upgrade 👑 (GitHub, Latent Colab, V-Diffusion Colab)

by Dango233 and multimodalart

Our Majesty Diffusion notebooks got plenty of updates - the most notable one is an improved upscaling schedule by Dango233. Check it out:

Another chance to sneak peek while @multimodalart is merging the code... pic.twitter.com/ie1GVEnrpd
— Dango233 (@dango233max) May 29, 2022

Text-to-video updates:

CogVideo hinted (GitHub)

by THUDM

CogVideo is a new text-to-video technique that was hinted on it's own GitHub repository. It seems to be capable to generate a few video clips from just text. From the creators of CogView and CogView2

CogVideo joins:
Nuwa (https://t.co/2B9oZvO854)
Video Diffusion (https://t.co/NY2fe0T03x)
Time Sensitive VQGAN (https://t.co/iL8CU82caQ)

As a direct text-to-video models. From those, only TATS have released their code so far. From past CogView experience, CogVideo may do it too
— multimodal ai art (@multimodalart) May 29, 2022

Learning AI Art:

AIAIArt course (GitHub, Discord)

AIArt is a free and open source AI art course by John Whitaker. There are synchronous classes for the next few Saturdays 4 PM UTC on Twitch. All previous classes stay recorded and available on Google Colabs on the GitHub link