How I use AI

I use AI in several parts of The Monk & The Witch, from building the website to organizing story notes, producing audio and working on prose. Its role differs in each part of the project. This page explains what I use it for, where it contributes to the finished work, and what I am changing.

Art

@PatriciaPtvArt, a graphic artist and Instagram influencer, has joined The Monk & The Witch. She will replace the project’s AI-generated artwork. She does not use AI in her creative process.

I used GPT-Image-2.6 Sunburst to create the current cover artwork.

Voices and audio

The narration and character voices are generated using Free and Open Source AI text-to-speech models.

I spent considerable time researching the available models, their source recordings and their license terms. I looked for voices built from recordings contributed or made available for legitimate sharing and reuse.

Respect for the people behind those recordings matters to me. I choose voices in good faith, based on the source information and documentation I can find, with the aim of avoiding models built from recordings taken or reused against the speakers’ wishes.

All the computation needed to generate these voices runs locally on my own computer. I care about open-source software and the freedom to run and maintain my own tools. This approach lets me produce the audio without depending on a paid external speech-generation service.

The models, source datasets, production tools and license references are documented in the detailed voice attribution notice.

Writing

AI is involved at several stages of my writing process: recording ideas, organizing the project, helping me learn, generating drafts and challenging my editorial decisions.

For those tasks I use Free and Open Source tools and models: Hermes and LongCat-2.5.

Story development, notes and organization

I develop my characters, world and story by talking through my ideas. Speech-to-text captures what I say, and I use OpenAI models to help turn those spoken thoughts into clean, organized notes in my knowledge base.

All of my note-taking and the organizational management of my books, chapters and draft structure go through this voice-based workflow. I describe ideas and structural changes aloud; the models help record and arrange them in my knowledge base.

I also use these models to assist with online research while developing the project.

Learning the craft

I have used different AI models over time to help me understand writing techniques, explore what makes a writing style, and put together guides for my own learning.

I still use AI alongside writing classes and advice from creators on YouTube as I develop my craft.

Drafting and rewriting

I use AI to generate prose drafts, which I listen to in my audio studio. Those drafts rarely survive in their initial form. I am usually dissatisfied with the writing and end up rewriting most of it, often from beginning to end.

I estimate that I typically rewrite around 90% of an AI-generated draft before I am happy with it. Sometimes I retain a description, a phrase or individual words. That percentage describes my experience of the process; it is not a measured figure for every chapter.

The prose I publish comes through that process of listening, choosing, rewriting and refining.

Editing and revision

Once I have written a chapter, I listen to it in my audio studio and revise it further. I rewrite passages, change sentences and adjust the narration myself. Listening is an important part of how I work through the text.

During this process, I use AI to capture revision notes and help me think through how to improve a passage.

When I know what I want changed, I sometimes give spoken instructions and ask AI to apply the revision directly to my text, particularly when I am away from my computer. This can include asking it to write the replacement wording. I direct the change through speech, and the result becomes part of the chapter’s continuing revision process.

Editorial feedback

I also use AI extensively to challenge my thinking during revision. I ask it to question passages and suggest changes, then work through whether those suggestions fit what I am trying to write.

This often involves long discussions and disagreements. I frequently reject its advice and make my own editorial choices. Sometimes, though, I think a suggestion is right and use it in the text.

There is no fixed rule for which suggestions I accept. Each decision depends on the particular passage and what I want it to achieve. It is a detailed, time-consuming part of my process.

Website and infrastructure

I use AI to write code and help me build and maintain monkwitch.com. That includes what people commonly call “vibe coding”: describing what I want to build, working through the implementation with AI, and refining the result.

I bring 15 years of experience as a full-time web developer and nine years as a site reliability engineer to this process. I have a deep background in building and operating websites and internet services, and I review a substantial amount of the code AI produces. I guide the implementation, assess the results using my engineering experience, and remain responsible for the software I publish.

For coding and infrastructure work, I currently use OpenAI models exclusively, mainly the GPT-6.x family, including Astra and Sol.

My aim is to move to open-source large language models as soon as I can. I am currently exploring how to make them work for my coding and infrastructure needs.

For my authorship and rights statement, see Rights. Technical credits and license references are available under AI voices and Attributions.