Review: An Illustrated Guide to AI Prompt Mastery by Jack Wylder

If the name sounds familiar, Jack Wylder does a lot of work with Larry Correia including producing Correia’s podcast. He’s recently produced a book which, unsurprisingly was promoted in former Puppy circles. That’s where I saw it but my interest wasn’t the connection to that particular circle of authors. Rather, I’ve been interested to see how independently published authors would start engaging with machine-learning art generation systems such as Midjourney and Dall-e for producing book covers.

Regular readers will know that I have done some of my own experiments in that area and for a deeper look, Ursula Vernon has done a number of Twitter threads showing her own experiments: https://twitter.com/UrsulaV/status/1555017878629486592

Getting a tool like Midjourney to make an interesting picture isn’t hard but what is difficult is getting these tools to make a specific kind of picture. The art/craft of generating pictures with these tools lies in experimenting with text prompts that provoke the machine to generate a picture. The syntax of these prompts varies between systems, for example, there is a detailed guide to Midjourney here https://midjourney.gitbook.io/docs/user-manual

An Illustrated Guide to AI Prompt Mastery attempts a system-agnostic approach to prompts. It doesn’t suggest a given system or discuss the syntax differences between systems. That is a sensible choice given that new systems are appearing regularly and the details of the syntax are better covered in their own documentation. The downside is that if you are expecting a kind of plug-and-play manual to AI-art syntax you’ll be disappointed.

Instead, the book works as a kind of catalogue of reference images. It’s only three chapters long but the first chapter is quite lengthy. Each page is a picture of a cat or a dog with the prompt used in the form of animal setting [in the style of X], for example, “beagle wearing a baseball cap [in the style of Julie Bell]”. Wylder does state that it will take several goes to get the examples of the quality he displays. Simply plugging in the prompt to a given tool won’t give you the picture on the page but that’s not the point of the book. Rather, Wylder is presenting a catalogue of prompt elements and using the images as reference points. In some cases that work well (eg cat [in the style of Tim Burton]) but in others less well, such as a picture in the style of Bloom County cartoonist Berke Breathed that is a great picture but not actually in the style of Berke Breathed, so not so great as a reference image. But, I’m picking holes, I quite liked this approach overall.

The second chapter has a broader range of animals and illustrates per picture a single art style such as engraving or expressionism. Again, you’ll need to look at the documentation of specific systems to see how best to apply a style in a prompt but the visual reference here is again useful.

The third chapter looks at how to make use of the art you generate to create better potential book covers. Wylder sensibly suggests generating multiple images separately and composing them together in software such as photoshop. By looking at a central subject, a middle ground and a background, Wylder shows how to make a decent-looking book cover that doesn’t look like it has just been spawned directly by an AI art tool.

If you are an author who wants to dive straight into using a specific tool to make covers for your books, this book is not a step-by-step guide. However, if you have already done some experiments with a given tool and already have some sense of its quirks and syntax, I think this book would be genuinely useful. It will be interesting to see how generated book covers will progress in these spaces over the coming years.