Every Picture a Story - What AI Art Can and Can't Do
There’s a little bit of everything in this newsletter - storytelling, what AI looks like illustrating children’s books, and some inspiration from the past.
Last week, a federal judge upheld a decision from the US Copyright Office that AI art isn’t copyrightable. The decision stressed that a human guiding hand was a “bedrock” requirement. That’s good news for human creators, but this isn’t likely to be the end of the ongoing battle.
How good is AI art?
Recently, IDEO designer began playing with ChatGPT and MidJourney to make a children’s book. He had heard that MidJourney had come out with a watercolor filter, and he trained it on the style of a revered children’s book illustrator of the past, Leonard Leslie Brooke, to see what it would come up with. He is currently selling his book in online bookstores. Here are some of his results:
On the one hand, these are pretty solid illustrations that do plausibly look like watercolor illustrations from an old children’s book, aside from some very thick straws. But the pigs don’t really have personality, do they? There are no dynamic poses in their bodies, no lively expressions on their faces. Would these illustrations have made you laugh, surprised you, or changed how you read the text? Great children’s books make children and parents linger over pictures, notice things they hadn’t before, and wonder.
Here’s the final pig in his brick house:
Now let’s look at the illustrator MidJourney was trained on, Leonard Leslie Brooke:
The dynamic curves of both the wolf’s body and the swirling straw add visual interest, while the pig’s expressively terrified face adds humor.
Here’s Brooke’s brick house pig:
Brooke wins hands down. It even raises questions - like why is he holding a spoon? Typical AI art may be beautiful and it may make you marvel at its technological achievements, but it’s a poor storyteller at best.
Here’s an example from one of my earlier drawings that also makes this point about storytelling. Captain Nemo (owl) has just surprised his prisoners with a feast where he brags about all of his inventions on the Nautilus. Nemo is getting carried away with himself bragging about all of his innovations. Professor Arronax is thoughtful but can’t resist thinking about how amazing Nemo is, while Ned the harpooner (weasel) thinly veils his contempt at being held captive, and Conseil, ever attracted by the latest advances, tries his paws at chopsticks and sushi even if he hasn’t a clue what Nemo is spouting about.
A closer look at Ned’s frown:
It’s like in a great film - every actor in a scene knows why they are there, what their story is, and how they are responding to their surroundings.
In art school, I remember when the idea of putting 25 things in a picture seemed overwhelming, with little energy left for thinking about worldbuilding or acting, but that’s what makes a world seem to come to life beyond the confines of the pages.
One last example from an old illustration of mine - teatime on a luxurious passenger liner. I hadn’t finished it, but then the story changed and I found it wasn’t necessary.
All the characters have a purpose. The raccoons may have been a middle-aged couple enjoying their cruise, the rabbits in the distance may have been an elderly aunt taking a trip with a younger niece, the badger waiter is a dapper-looking fellow looking more trim and poised than your average European badger, and in the foreground, two foxes having their moment where a proposal may be in the making. This was a setup for the next crash scene, where everything gets tossed up in the air after the Nautilus rams the Scotia.
If you think back to your favorite children’s books, the stories and pictures probably dissolved seamlessly into each other. Think about the feelings these stories gave you. It’s a good thing that we can treasure what humans can do.
Please share my blog with anyone who might enjoy it. The more the merrier and thanks! Krister
Any thoughts about AI art yourself?