What’s the purpose of key art in a blog post or newsletter? Really, it’s just a splash of color. Something to settle your gaze for a moment; starting point, foyer. Yes, that’s the right image: an inviting door, rather than a wall of text.
At OpenAI’s blog, the designers have given up and just use gradient blobs.
I think both approaches are Fine, Actually. The only option that’s not acceptable is AI slop, because, if the image doesn’t matter anyway (and it doesn’t), why not use something real? There exists a vast trove of visual art in the public domain; it has never been easier to browse and search; and all of this art was made by someone, who lived and worked and dreamed. Some of it is so, so strange — in a way that AI slop never is, maybe cannot be.
I’ve used public domain art in my newsletter for years. Here are some favorites:
The art almost never reflects “the content” directly, though I do tend to go find it after drafting the newsletter, so one hopes there’s subconscious resonance. Sometimes, I add a short paragraph about the artist at the end of the newsletter; other times, I don’t. The point isn’t to be didactic or museum-ish; the art is really, truly just a splash of color … that happens to summon, for a moment, to thousands of screens across the world, the vision of an artist long dead.
And if it catches the eye of a reader, and they follow a link to investigate, there is something waiting for them. A whole life.
Where can you find public domain art? Chase McCoy’s Museo is a great starting point. The Cleveland Museum of Art’s online collection is fabulous, and they make it super easy to download high-resolution images. The Met is a treasure trove, obviously — especially for ancient images and artifacts. The Rikjsmuseum is packed with wonders, though you do have to make an account to download images.
Finally, I wish Artvee linked to its sources, but there’s no denying it is a vast, surprising archive—and it’s important to insist that the power of public domain art is that you can do anything you want with it, so their cruddy consolidation is, in its way, an affirmation of the whole system.
2.
Rex Sorgatz’s treatise on key art for movies and TV shows is part of my internet canon: a gimlet eye turned towards something you maybe didn’t realize was a genre, a craft.
Near the end, Rex contrasts the finesse of good key art with Netflix’s algorithmic presentation:
On Netflix, each user now sees different key art, based on their user activities and ultimately their demographics. Which helps explain why you and your partner can never agree about what to watch. You are literally presented with different options.
To illustrate, I loaded up my wife’s Netflix profile next to mine. We got some of the same recommendations, but the artwork makes the shows look completely different:
Rex goes on to complain that the algorithmically-tuned key art is ugly … but/and this was 2021, and I wonder how the state of the art has advanced in the years since?
Is there any chance that Netflix is now using AI models to guide and evolve these pivotal graphics? (Yes, there is a good chance.) What about using AI image models to actually generate them, and if the AI-generated tiles do not exactly represent what happens in the movie or show, well, the same has been true of movie posters for decades? (Here, too, there is definitely a chance!)
Naturally I’m now imagining that Netflix might insert into the home screen, every so often, key art for a movie that doesn’t yet exist. When a user selects that tile, the system meekly reports, “Oops, this content is not available”—not technically a lie — meanwhile tallying a vote for its production. Above some threshold of performance, a Netflix executive (or maybe it’s all software at this point) actually greenlights the movie, fulfilling the prophecy.