Hello, my name is Florrie, and I am a so-called “AI-specialist”. The other day, my musician friend June bragged that she would not need to pay anymore to design her album covers. She would instead use the text-to-image diffusion model Stable Diffusion.

Stable Diffusion is an AI system that creates art from text. For example, here is a recent picture of me and CC:

In Stable Diffusion, after typing “Florrie getting conjunctivitis from false eyelashes”, you get this image:

Pretty cool huh?

Stable Diffusion is based on deep-learning, learning from a large amount of examples and picking up patterns. This is perfect for June, who sometimes releases several albums a day, and therefore has plenty of examples to train the AI.

Even Wool started using Stable Diffusion, as its source code is fully available. Below is his AI-generated dream house:

Based on real-life pictures of June including this one:

the AI was trained to generate an image featuring June and friends. It will be the cover for June’s upcoming album, the most personal she ever released, “Je. Je suis June… The Protrusive Artiste”. Here is an excerpt:

As June is quite prolific, and as I love everything AI-related, it was natural for us to work together on finding the best next covers for her. Together we settled for the covers for:

• "Pigeon Shin Chan"
• "pigeonzilla"
• "If You're Happy And You Know It, Flap Your Hands"

As you may have noticed, Stable Diffusion has limitations when to comes to generating limbs. It will likely need more training data to correct this issue.

All and all, the results are promising, and it will be an opportunity for many artists to do some vocational retraining. We are more than ever in need of video editors, interior designers and embalmers, and hopefully AI art generators will greatly serve this purpose.

(The opinions expressed here are those of the authors. They do not reflect the opinions or views of the Government or the Cosmoose Gang.)