The cover for Damien Pollard and Edward Bowen's new collection Film Exhibition: The Italian Context, which appears in the Moving Image series in early 2024, is a superb photograph by Fabrizio Fioravanti of Rome's Cinema Induno.

And we'll get to that, but I was curious to see what the new machine-learning model Stable Diffusion might design instead. There's been talk of AI putting magazine and newspaper illustrators out of a job: maybe book cover designers will go the same way? Well... not this year, but the time may come. This is the image invented by Stable Diffusion (I used the M1-based MacOS port Diffusion Bee) in response just to the title "Film Exhibition: The Italian Context":

Diffusion result

It takes a moment to realise that this building does not exist, and that none of those people – if they even are people – exist either. The roof... does not quite look supported by the walls. What exactly is the churning mass in the middle of the space? Why are the faces deformed, squished, elongated?

Another try? This time the prompting text was "Italian Film Exhibition":

Diffusion result

Hmm: maybe. "Italian Film Festival" produced something of a Dali-esque smearing of faces and dresses:

Diffusion result

I quite liked the results of "Italian Cinema", though:

Diffusion result

For the time being, though... perhaps we should stick to human artists. The real, not diffused, cover, with its actual, not-hallucinated-by-a-neural-network, cinema on it:

cover of Film Exhibition

full news feed • subscribe via RSS