Where I stand on AI

I’d rather be straight about this than wave a slogan.

The short version: I don’t use AI to compose music. I never have, and I don’t plan to. Every theme, every melody, every arrangement on this site was written by a human being making decisions about a story. That’s the work, and that’s the point.

Everything below is the longer, honest version — because “no AI, ever” would be a tidier line than I can truthfully deliver. My catalogue has a few brushes with AI-adjacent tools from earlier in the journey, and a few places where I quietly use AI today. I’d rather show you the map than pretend the territory is simpler than it is.

The music itself

My compositions are 100% human.

• I write themes on piano and guitar, then orchestrate them.
• I record live players whenever the project’s budget allows. The rest of the time I perform virtual instruments myself.
• I do not use AI music generators. Not for sketches, not for “ideas”, not for production passes.
• I do not feed prompts to a model and arrange what comes out.

This will not change. Writing the music is the part of the work I most want to keep, and the part I find the most fundamentally human. I find the idea of an AI writing the music for me self-defeating — there’s nothing left to do that’s mine.

A few honest footnotes

A couple of releases on this site predate the current AI conversation, or were made while I was still figuring out what AI in the arts meant. I’ve chosen to leave them documented rather than rewrite history.

Synth-V vocals on the Minstrels of Mirth albums

The Minstrels of Mirth series uses Synthesizer V for the sung vocal parts. Synth-V is a voice-synthesis tool: I composed the melodies, wrote the lyrics, programmed every phoneme and inflection, performed the parts, and mixed the result. The “AI” in Synth-V is a vocal-modelling engine, not a generator that wrote anything. The decisions are mine and the songs are mine. Without a budget for a soloist and a string of vocalists, this was how I could get the music I heard in my head onto a record.

I treat Synth-V the same way I treat any other virtual instrument — Kontakt libraries, sampled choirs, modelled brass. It’s a tool that lets one person realise a larger ensemble. I’d still rather work with live singers when the budget allows.

Historical AI cover art

Some older covers and blog images on this site were generated with image models during the period when those tools first became broadly available. I used them while I was learning — and then learned more.

I’ve since stopped commissioning AI-generated imagery and prefer commissioning human illustrators or using stock photography when needed. Older posts and album art from that earlier period are still here. I’d rather leave the trail visible than quietly rewrite the back catalogue.

Where I use AI today

I’d be lying if I said I avoided AI everywhere. The honest list:

• Marketing copy and emails. I sometimes use AI assistants to draft newsletter copy, social posts, or to tighten my own writing before I publish it. Final words and edits are mine.
• Site copy. Some pages on this site were drafted with AI assistance and then edited by me. The opinions, decisions, and the writing voice are mine — but if you read closely you may catch the assist.
• Assisted Development. I worked as a web developer for 4 years of my life, nowadays I’ve been using AI to assist me during code-related tasks. Or take care of highly tedious tasks.

I don’t use AI in the music. I don’t use it to generate cover art any more. I don’t use it to fake reviews or fabricate quotes.

What I think about it

I’m not anti-AI on principle. I’m against using it for the parts of the work I want to keep doing as a human — and composition is the centre of that for me.

I don’t say “no AI, ever” because I don’t know what the next decade looks like, and I’d rather be honest than perform a stance. What I can commit to:

• The music will be human-composed. Always.
• I will keep this page updated if the way I use AI changes meaningfully.
• If you license a track or commission a score, you can trust that no AI generator was involved in the writing of it.

What about AI training on this site?

The music itself isn’t on this domain — every audio file lives on a separate CDN behind signed URLs and access controls. Training crawlers can’t reach it, and they shouldn’t.

The text, blog posts, song descriptions, and cover art on the rest of the site — that I leave open to AI crawlers. If a model is going to recommend royalty-free fantasy music to someone, I’d rather it know I exist than not. There’s no contradiction with the position above: AI tools learning that I write fantasy music for tabletop campaigns is fine. AI tools learning how to compose fantasy music from my recordings is not. The structure of the site enforces the difference.

If you have questions about a specific album, track, or how something was made, I’m happy to answer — [email protected] or Discord.

— Ivan