Learn to make artificial intelligence (AI) images

NB: The content of this page has been superseded; Craiyon is not part of our Learning Together resource repertoire.

This is fiction

The image at the top of this page shows the site of a fictional 50+ Digital venue, before we moved in. The image below is a composite of paintings made at the fictional inaugural launch party. Further down this page, you will find a portrait gallery of the first 24 fictional learners to arrive at the fictional venue.

5 separate panels, each containing a small square dance scene - all in a vibrant imaginary forest glade, and all obviously artificial

All the images were made with Craiyon, an artificial intelligence image generator. It’s a free online tool that is relatively easy for beginners who would like to explore the basic features of AI by actually doing something with it.

To create the venue and record everything that happens in it, all we had to do is describe – in plain, ordinary text – what we wanted. We thought you would appreciate a glade in a mythical tropical forest where everything is over the top – and all the learners are either using a digital device or partying, or both.

Using Craiyon – an AI portrait gallery

Craiyon is easy

Craiyon does not make the best AI images, and it is very slow – but it’s also free, relatively easy, and does not require a login.

What’s in the gallery, and what to look for

24 images of fictional learners at the fictional 50+ Digital Centre in a mythical forest. They are all supposed to be older learners using a smartphone or a laptop. Some have a passing resemblance to non-artificial celebrities.

No older people were harmed in the making of this gallery.

How to use the gallery

Click on any image to see a larger and more detailed version. To return to the gallery view, click anywhere outside the larger image or the small x near the top right inside the image.

The larger versions are slides in a looping slideshow (look for the < and > symbols near the left and right edges of the larger image).

Your opinions please!

Some of the images are intended to be photorealistic, others to look like paintings. How successful was that?

Which do you think are more convincing – the small thumbnail images or the larger versions?

Did you count the fingers on their hands? (Look again at the ‘Bob Marley’ and ‘David Attenborough’ look-alikes).

Why are the flaws so obvious when we look closely?

Perhaps it’s because we are looking closely. As humans, we can’t not notice the defects and deformities – even if we are impressed by the overall achievement.

Like other generative AI systems, the Craiyon model does not know what it is doing. It does not know how accurate it is, because it cannot know anything.

All the images it makes are based on fragments of information taken from a vast digital warehouse of images scraped from the internet – and none of it is original artwork. How that happens is way beyond the scope of Learning Together.

Make your own AI images with Craiyon

The reason we are starting with Craiyon is because it is easy. There are AI image generators that are much better, but require a sign-up process.

Visit craiyon.com

Exercise 1 – imagine Liz Truss at the digital drop-in

In the box that asks ‘What do you want to generate?’, enter this prompt (copy and paste is the easiest way) –

Happy Barbie, resembling “Liz Truss”, using a smartphone, in a lush rainforest glade, Pre-Raphaelite style, long flowing satin gown, silver hair, ethereal, saturated, photorealistic, smooth, concept art

Specify in turn each of these models: Art, Drawing, Photo and None.

Each model will give you 9 images, but don’t expect all of them to be useful.

Ignore the ads — they are the reason you are not paying money to use Craiyon.

Exercise 2 – illustrate a story

In the box that asks ‘What do you want to generate?’, enter this prompt – using models Art and Drawing

A puppy wearing a blue hat, in a small boat, floating on a calm green sea, watercolour painting

The second of the two images below is our result from a similar prompt.

Next – try something completely different, using your own experimental prompts — the only way to learn.

AI prompts in Craiyon

Like all other AI systems, your Craiyon results depend on your knowledge of how to build prompts. Read more about that on the Craiyon blog – craiyon.com/blog

How useful is Craiyon?

It’s not designed to be useful. The developers have built it as a consumer game, not a professional tool.

It doesn’t generate art, and it won’t help you become an artist. There is nothing artistic about the process nor the result.

Craiyon is marketed as a consumer product. You can see an example of that sort of use on our blog post Xmas Lunch Party at Mildmay Community Centre (the image has additional text to make it a Christmas card).

However – there may be a role for generative AI imagery in concept artwork and storyboarding. But Craiyon (as it is now) is not sophisticated enough for that.

Post a comment, reply or question

This site uses Akismet to reduce spam. Learn how your comment data is processed.