Visualizing Characters

Midge Ure in 1985.

When I created Roland Axam, more than 40 years ago, I had a clear picture of how I wanted him to look. My favourite band at the time was UK new-wave band Ultravox. Around that time, front man Midge Ure had started releasing solo albums, and the cover for his first one, The Gift, featured Ure in a black-and-white, slightly noir head shot.

Roland Axam, in my mind, didn't look exactly like Ure but close enough to win a look-alike contest. Roland had a more-defined jawline and eyes that always looked half-closed, like he was either in deep thought or was sizing you up.

If you ever looked at Roland, you would never know what he was thinking. And that was perfect for the short stories and trilogy that I wrote, in the late 80s, where Axam was a spy.

Me, in 1998.

By the time I wrote Songsaengnim: A Korea Diary, Roland looked like a cross between Midge Ure and myself, especially how I looked at the end of my first year whilst living in South Korea. I often visualized Ure, on the cover of The Gift, and a photo of me in a Seoul market, and I mashed them up in my head to create an image, but that image was never clear to me.

When I write about other characters, however, I don't always see a face.

When I started writing Dark Water, a few months ago, I didn't have images for most of the characters. When I created the victim that is discovered in the Rideau River, I just imagined a woman in her late 20s or early 30s, with long, red hair.

Once she was extracted from the water, I kept picturing one face as I wrote about her. And I'm a bit disturbed to have my image be that of Ottawa journalist, Rachel Gilmore.

I say 'disturbed' because while she fits the profile of my victim, I would never want to see any harm come to Gilmore. She's great at her job and I enjoy following her on Bluesky and YouTube.

So far, I've been unable to clearly visualize my two inspectors, Mickey Calloway and Erin Hayes. In fact, a couple of weeks ago, when I asked ChatGPT to create a book cover that included these characters, I purposely requested that the detectives faced away.

The AI tool was actually pretty good at visualizing my description of the detective duo.

But this week, as I was reading through my rough manuscript, trying to flesh out the characters more, I still couldn't give Calloway or Hayes a face. So I finally asked ChatGPT to do it.

Image: ChatGPT

The first attempt was pretty good and I liked the face that was given to Calloway. I think he could be 'drawn' taller and more imposing (he's a solid 2 metres and broad-shouldered).

And while the image of Hayes is good, I think she appears in the image older than she is in the book. She's supposed to be 28 and, as Calloway put it, is "ridiculously beautiful." I mean, the woman in the ChatGPT image is pretty but not ridiculously so, IMHO.

I then asked the tool to make her prettier and younger, reminding ChatGPT that she's supposed to look 28. I think the woman in the first image looks mid-30s.

Here's what it gave me in its second attempt:

Image: ChatGPT

Closer. I prefer the Calloway rendering in the first image but Hayes looks closer to her age this time.

As I've written her, Inspector Hayes is half a metre shorter than Calloway. In the image that I had ChatGPT create where the detectives are faced away, I asked it to change the image so that she was 75 percent shorter than Calloway, and the tool gave her the legs of a child but kept the proportions above her legs the same.

Talk about ridiculous. Obviously, I didn't keep that image.

I'm not going to mess with the characters' height in this image. It's not important. I just wanted to create some faces to give me a better visualization for when I write the story, and I think these images help accomplish this objective.

Unlike how ChatGPT visualized Roland Axam when it merged our photos.

Happy Tuesday!

ChatGPT's first interpretation of a merge of me and Midge Ure, as Roland Axam. Hmm...