My case: building a one-person AI music label — part 1, from zero to an album on streaming

I wrote the honest breakdown of whether a solo can make money with AI music — the artist path versus the studio path, the economics, the spam purges. Fair question back: would I run it myself? I am. This is the live log of that experiment: a one-person AI music label, built the same way I build everything else — structure first, minimal ritual, real results or an honestly logged failure.

Part 1 — what’s built (the setup stage)

• The catalog. Produced with AI music tools on a paid plan — the paid tier is what carries commercial rights to the output — with my own concepts and lyrics as the creative layer, and every draft archived as authorship proof. Multiple genres and artist personas rather than one act: this is deliberately the studio model — breadth as inventory, not one name chasing luck. (I’m keeping persona names out of this log for now; results first, branding later.)
• An album, assembled and shipped. The first full album (blues) went through a standard digital distributor to Spotify and the other stores — real releases, real catalog, not files in a folder.
• The legal spine. My company owns the rights; the label is its imprint; songwriter credits use my real name. Same clean-setup discipline as any business: the asset is ownable and defensible before it earns.
• The YouTube layer. Channels built around long-form mood compilations — because watch-hours are YouTube’s real currency, and streaming’s fractions-of-a-cent make Spotify presence, not income. There’s a clock on this: YouTube’s partner thresholds are set to roughly double in early 2027, so channels started now have a window to qualify under the current bar.
• Two royalty pipes, one deferred. Master royalties flow from the distributor now; publishing/songwriter royalties need a collecting-society or publishing-admin registration that only pays off at scale — so it’s deliberately deferred until a real trigger (meaningful streams, or someone covering a track). Minimum machinery, added only when the numbers justify it.

Why this maps to everything else on this site

Strip the music out and it’s the same solo playbook: one owned asset (the catalog), boring compounding distribution (long-plays ≈ SEO), structure before revenue, and no invented numbers. It’s the studio path from the AI-music breakdown executed for real — and the same “craft, not spam” line: a focused catalog with human intent, not 500 generic uploads that platforms delete.

The log — results as they land

• 2026-08 (part 1, this post): catalog in production across several genres and personas; first full album (blues) live on streaming via a distributor; legal structure done (company → label imprint → personas); YouTube long-play channels launched and warming. Revenue so far: nothing worth reporting — as expected at this stage. Next checkpoints: first distributor royalty reports, channel watch-hour trajectory.

• 2026-09-19: the audience layer got its own log — the channels that grew out of this catalog, the analyser I built to understand them, and the first weeks’ honest results: launching YouTube channels solo, part 1.

Updates get appended here with dates as real numbers arrive — wins and failures both.

The step-by-step version: how to make and release music alone.

Part of the personal log. The framework this case tests: can a solo make money with AI music?

The thinking this experiment belongs to: living on the border of the synthesised world.

The recipes behind this log, for anyone who wants to run them: make a track · publish yourself · all of them in Do it yourself.