How I'm Using GenAI to Help Me Write a Cozy Mystery
Will be mostly enjoyed by those of you who like reading/watching mysteries, and writing!
Note:
I’ve cut down most of the exchanges with ChatGPT. They’re all in italics if you want to skip them. If however, you’re curious as to what the whole conversations look like, let me know and I’ll share them with you.
It’s taking me ages to write my first mystery, but it’s ok, because it’s a lot of fun.
A cozy mystery is a crime story with no gore and no sex. There are many other “tropes” and unwritten rules, like the fact that it usually takes place within a tight community.
Mine takes place in a recording studio, in Soho and the protagonist is a voiceover artist, called Mia. (Write what you know!)
I’m almost done with my fifth “first draft”, thanks to spending two weeks focusing on the project. As some of you know, I’ve been using ChatGPT and Claude to get feedback, but I’ve also used them to help me with scenes. Here are some ways in which I’m using GenAI, mainly GPT-4.
Feasibility of the Crime
(I felt a bit nervous writing the above title. I don’t want anyone who’s looking for ideas to murder others to land here… And that’s also an issue when interacting with ChatGPT - you need to be careful that it doesn’t think you’re planning some sort of crime, or else, it will refuse to reply to your questions.)
I’ve always known that the first victim in the mystery was a sound engineer, and that he died in a studio. I first thought he had been poisoned, and found dead at his sound desk. But then, I attended a voiceover session with my friend Fran, who has a great imagination and reads a lot, and I shared with him that I wasn’t 100% happy with this choice of murder. We got talking and came up with a change of scenario: the engineer would be murdered by being hit on the head.
But how come no-one heard him scream, or heard strange noises?
It happened in the soundproof booth.
As you can see, and as I’m sure you already know, writers get their ideas from all sorts of places, including other people and their friends. So that’s why I have no problem in asking a bot to help me with my writing - it’s one more source.
Having decided that the engineer would be killed by someone hitting him with a cue light on the head, I wondered how realistic this was. Here are some of the questions I asked ChatGPT. (I’m not including all the answers here, so leave me a comment or send me a message if you would like to see the full exchange.)
Q1 I’m writing a cozy mystery. The murder takes place in the booth of a sound studio. The idea is that the engineer has been killed by someone, in a fit of rage, who takes a cue light and hits him across the head. I’m unsure of how realistic this is. My questions to you: is this realistic? What would be the effect on the surroundings eg would there be blood splatter etc
Q2 What materials are cue lights made of and which would be the heaviest ones?
CHATGPT
[…] However, even these robust cue lights are not typically so heavy that they would be the first choice as a murder weapon. For your story to be believable, you might need to emphasize that the cue light was used out of immediate convenience, in the heat of the moment, rather than being a premeditated choice.
ME
Yes it will be obvious that the cue light is used in the heat of the moment, but it might not be believable. Unless the murder victim has some condition that kills him, like a heart attack?
Q3 Could asphyxia be caused by falling head down on the carpet?
CHATGPT
Yes, it's possible, though not particularly common. This is known as positional asphyxia, which can occur when someone's position prevents them from breathing adequately.
ME
I think I’m going to go with him being hit unconscious by the cue light, and then asphyxiated by being on the carpet for awhile. Do you think a coroner would be able to come up with this explanation?
CHATGPT
A trained medical examiner or coroner could potentially come to such a conclusion. The process would likely involve several steps: [Lists steps]
Q4 How long would something like this take in real life? And how long would it be acceptable to say it takes in a cozy mystery?
Q5 My protagonist, is not in the police, and is not a relative of the victim. But she has access to someone in the studio who is willing to give her information. I’m thinking of a believable way for her to get the information.
As well as helping me get closer to reality in a work of fiction in a world which, to be honest, I know little about, asking these questions helps me to think about the scene in detail.
Describing the Scene
Descriptions are not my forte. If a book has a set of paragraphs describing a setting, I skip them. I’ve also never had to describe something in detail, in order to help an amateur detective figure out how a crime was committed!
So that’s something I sought help with too.
ME
I'm having trouble describing the microphone. It sits on a microphone arm, and instead of hovering over the table, as normal, it looks like someone swung it across and now it's hovering over the floor. Can you help with that?
CHATGPT
Of course, here's one way to describe the situation:
The microphone, usually poised above the table like a diligent sentinel, now hung askew. Its arm was stretched out as though someone had forcefully swung it aside, and the microphone itself hovered over the floor, a silent witness to the grim scene. The change in its usual position was as disconcerting as the lifeless body beneath it.
You can see why I don’t use ChatGPT to generate text for fiction (as opposed to non-fiction, where it does a better job)! However, if you take the commentary out, the description is clear.
Identifying the Cause of Death
Many cozy mysteries have a second murder (this victim is usually the main suspect of the first crime!), and so does mine. He’s also not a very nice character, which also makes it easier to kill him.
This time, I wanted the victim to be poisoned in the studio, but I wasn’t sure how.
ME
I need a few alternatives on how someone could have poisoned John. I like the vial (NOTE: I’d asked the bot to give me some ideas), can you give me other similar ideas?
Also, I'm thinking that another alternative could be for the poison to be on his toast. Maybe he could have had something with a very strong taste, like marmite? What do you think?
CHATGPT
Certainly! Here are some alternative methods on how someone could have poisoned John:
1. **Teabag Tampering**: Maybe John always had a specific brand or type of tea, which most others didn't drink. The poisoner could have carefully laced his specific teabags with a toxin that dissolves in hot water. It might be a slow-acting poison, so it wouldn’t activate immediately after his first sip but sometime later.
2. **Sugar or Sweetener Sachet**: John might've had a routine of adding sugar or artificial sweetener to his tea. A pre-poisoned sachet could have been prepared and left in the bowl for him to use unsuspectingly.
3. **Lip Balm or Chapstick**: If John used lip balm, especially in the winter months, it could have been swapped out for a poisoned one. This way, each time he applied it, he ingested a small dose, which accumulated over time.
4. **Stirring Stick**: If John often used a wooden stick or spoon to stir his drink, a specific one could have been pre-treated with a fast-acting poison. Once he stirred his drink with it, the poison would mix with the beverage.
For the toast idea:
5. **Marmite or Spread**: Given that Marmite has a strong taste, it's a good cover for something off-tasting. A poison that pairs well with salty or umami flavors could have been mixed into John's Marmite jar. Since not everyone likes Marmite, it would mostly target John, especially if he’s known to enjoy it regularly. If the poison was colorless or similar in shade to Marmite, it would go undetected.
6. **Butter or Margarine**: If John had his own personal tub or stick of butter or margarine (perhaps a specific dietary or brand preference), it could have been tampered with. A toxin mixed into the butter would easily go unnoticed, especially if spread on flavorful toast.
7. **Special Toast**: If John had a preference for a particular type of bread, like gluten-free or a special grain type, the slices could have been pre-treated before being toasted, especially if the poison can withstand the heat of toasting without losing its potency.
You can see how this is highly addictive. It’s great fun just to read through all of these suggestions!
I’ve also used it for more practical help, like asking what were some popular music tracks in 2018, when the novel is set, what’s a pop song that’s had many covers and: “I'm thinking of setting the novel around the time of the Brexit referendum. Would there have been less police available for day to day crime, or would crimes like this murder have taken priority anyway?”
I could turn to books to find out most of this information. Indeed, I did read through “Murder Isn’t Easy: the Forensics of Agatha Christie by Carla Valentine”, for example, but to have someone you can bounce ideas off during your writing session is very useful - as it’s during these that most of my ideas come to me.
Feasibility of the Story
I know that not everything the bot comes up with will be true, but it’s not too big a deal, as I’m writing a cozy mystery. And that means that events don’t need to be as realistic as in other mysteries or novels. However, there is a limit to how much you can push the reader…
Here are some exchanges I’ve had with Claude:
ME
Would it be too much of a coincidence that the text that Bea sees on Rowan's phone has the address for the studio?
(NOTE: This is part of a subplot. Bea is the protagonist’s friend, outside the voiceover world, and Rowan is her boyfriend.)
CLAUDE
Using the address in the text message on Rowan's phone could be a bit too coincidental, especially if there hasn't been any prior connection established between Rowan and the recording studio. Coincidences can sometimes strain believability and make the story feel contrived if not handled carefully.
While ChatGPT was more encouraging:
CHATGPT
In a tightly woven mystery narrative, coincidences can sometimes stretch the bounds of believability, potentially taking readers out of the story. However, if handled with care, what might initially seem like a coincidence can be a pivotal plot point that adds depth and intrigue to the story.
I’m also worried that the police officer in charge of the investigation is sharing too much information with the amateur sleuth, but, how else is Mia going to get all the information she needs to solve the murders? I shared my concern with ChatGPT, who answered:
In the context of a cozy mystery, the interaction between Mia and DS Baker is engaging and serves the narrative well, but it does stretch the boundaries of how much a detective might realistically divulge about an ongoing investigation. Police officers, especially detectives, are typically cautious about sharing details of an open case to ensure the integrity of the investigation and to avoid influencing or jeopardizing it in any way.
A Way Into Your Subconscious?
To be honest, a lot of the answers I get from GenAI when I’m writing, either fiction or non-fiction, are disappointing.
But almost always, I solve the problem I had or come up with a new idea after I’ve asked the bot a question.
Maybe what’s happening is that:
1) In order to ask the question, you have to identify the problem, or next step. Asking the bot gives you that opportunity.
2) In realising that their answer is not the right one, you get closer to the answer you want.
3) Talking to the bot takes the focus away from you, and frees up your subconscious, which is where all the fun stuff lies.
Pretty much like writing here.
Every time I share my process with you, I discover something new about it. Let me know whether there’s anything else you’d like to know about writing with GenAI, or any other topic, around Spiralling Creativity.
Thank you so much for reading!
PS All my writing in this blog is my own, that’s why it sometimes reads as stream of consciousness - it is! (Kind of…)