I have been writing blogs for over twenty years. Fishing trips, cooking, travel, technical notes, whatever I was thinking about that week — across several sites it adds up to a genuinely large pile of text.
About six months ago it occurred to me that some of it might work as lyrics. I tried one, ran it through Suno, and released it. Since then the loop has been: pull a scene out of an old post, rewrite it as lyrics, generate the track in Suno, distribute through SoundOn. The songs are on Spotify. Some of them get used on TikTok.
This is the operational log — what the workflow actually looks like, what came out better than expected, and the problem that turned out to be the real one.
Originally published in Japanese on February 15, 2026. This English version was written in September 2026.
The rules changed on September 3, 2026
I checked all of this before publishing the English version, and the timing is awkward enough that it belongs at the top rather than in a footnote. Suno’s revised terms took effect on September 3, 2026 — four days before this went up — and they change the part of this workflow that matters most.
- Releasing a track now consumes a download allowance. Under the new terms a song sitting in your library may not be commercially exploited — released, sold or licensed — until you have spent one of your monthly downloads on it. Pro gets 20 a month, Premier 60. This is the single biggest change: the old mental model of “pay for a plan, generate and release freely” is no longer how it works, and a release cadence now has a hard monthly ceiling.
- The free plan cannot be used for this at all. On the free tier Suno retains ownership of the output and use is non-commercial only — and upgrading later does not, by default, grant rights retroactively over what you already generated. If you intend to release, be on a paid plan before you generate.
- Suno now reserves the right to watermark and fingerprint output. The clause covers “a fingerprint, watermark, or metadata indicating the applicable service tier of an Output and whether such Output was a permitted Download”, and separately forbids removing or obscuring it. That is a reservation rather than a statement that every track already carries one — but plan as though tracks will be identifiable as Suno-generated downstream.
- Your distributor and the destination stores increasingly want disclosure. Spotify takes AI credits through the DDEX standard and Apple Music has its own AI tag; both are optional today rather than required, and Spotify says it does not down-rank music for being AI-assisted. What does get tracks pulled is a false disclosure combined with impersonation or spam. So declare it, and declare it accurately.
Primary sources, both current: Suno on rights and ownership and Suno on distributing songs to Spotify and similar platforms.
And the lawsuits, since you already know about them
English-speaking readers will come to this with the label litigation in mind, so it would be strange to write around it. Where it stands as of September 2026: Warner Music settled and signed a licensing deal in November 2025, and BMG followed with a global license in August 2026. Universal and Sony are still litigating, with fact discovery closing on September 30, 2026. A Munich court has ruled against Suno in the German GEMA case.
Suno has also said that when its licensed models arrive — expected later in 2026 — the existing models will be retired. So the tool that made these tracks is on a stated end-of-life path, which is a strange thing to be able to say about the instrument you recorded on.
None of that is settled enough for me to tell anyone what the right call is. I am describing what I did, on the plan I had, at the time I did it. That is not legal advice and it is not a guarantee about your account.
Why an archive is unusually good raw material
I used to write songs — badly, alone in a room, with almost no band involvement. The fact that I stopped tells you how that went. Writing prose is the part that stuck.
What I did not expect is how well the two connect. The problem with sitting down to write lyrics from nothing is that you have no material — you are inventing a scene and a feeling simultaneously. An old blog post has already solved the material problem. The scene is specific, the details are real, and you were emotionally present when you wrote it. Turning that into lyrics is compression and rhythm work, not invention.
The first attempt was a post about herabuna fishing — Japanese crucian carp, fished in winter, at night. A quiet water surface, the red glow of an electric float, cold. Not an obvious subject for a song. But I had written about it repeatedly, so I knew exactly what the scene was.
I restructured it into verses, handed it to Suno, and got a real song back. That first one is genuinely startling.
What generation is actually like after the first time
The novelty wears off in about a day and the real character of the tool shows up. It does not do what you asked.
The genre drifts. The tempo is too fast. The vocal timbre does not match the world of the lyrics at all. You had a mood in your head and what came back is confidently something else.
And then, unpredictably, it is more interesting than what you had in mind. Chord movements and structural turns I would never have written, because I do not think that way.
The work is entirely in the gap between those two — iterating between “not what I wanted” and “better than what I wanted”, many times per track. What you end up learning is prompt craft: how to specify a genre, how to define a persona and keep it consistent across releases, how to describe a vocal quality in words that actually move the output. I lost count of how many generations went into the first track before I stopped, and I still would not call it exactly what I had imagined.
Expect to discard a lot. The generation is cheap; taste is the bottleneck.
A practical note on the tool itself: the current model at the time of writing is v5.5, released in March 2026, and it moved fast enough underneath me that tracks made six months apart do not sound like they came from the same instrument. Lyrics are pasted in directly, with a 5,000-character limit, and v5.5 added explicit duration control. Extend, cover, persona and stem separation all exist. If you are reading an older tutorial, assume the interface it describes is gone.
Releasing through SoundOn
I distribute via SoundOn, which puts tracks onto Spotify, YouTube Music and the other major services.
Not everything gets through. I have had submissions rejected, which was useful in the way that rejections usually are — it forces you to read the requirements properly instead of assuming.
The first release landing was a bigger moment than I expected. I drove to a fishing spot playing my own track in the car, which is exactly as ridiculous as it sounds. Text I wrote years ago, now sitting in a streaming catalog as audio.
The genuinely surprising part came later. SoundOn is operated by TikTok’s parent company, so tracks flow into TikTok’s music library with very little friction. I do not use TikTok. My songs are being used on TikTok anyway.
That is a distribution mechanism I have no equivalent of in blogging. Text reaches people who read; a fifteen-second clip reaches people who would never open an article about winter carp fishing. Same source material, completely different audience.
The actual tracks
Two projects, deliberately different.
odeco
The main one. Odeko is Japanese angling slang for coming home with nothing — a blank session. It seemed like the right name. Site: odeco.life.
- “hera” — the first odeco track and the first song I ever made with AI, built from the herabuna fishing posts. I did not want anything showy, and getting to something that restrained took a lot of attempts.
- 「夜釣り」(“Night Fishing”) — from posts about night fishing for conger eel in the rainy season, and a sub-zero winter night on the Kurobe River.
- “Emotional Flow” — not fishing. This one comes from a post about arriving early for drinks with a boss and waiting alone in a nearby park.
algohush
A second project with a different register — where odeco is calm, this one is psychedelic. Site: algohush.live.
- “Atamanurakkawa” — roughly, “the back side of your head”. Drawn from an introspective post in a style I would not write today. Watching that kind of writing come back as a song is a strange and specific pleasure.
Splitting into two artists was one of the better decisions. A single project holding both quiet fishing songs and psychedelia would have no identity at all, and identity is most of what makes a catalog findable.
Six months in: what works
- A text archive converts into a music catalogue. Years of writing that had exhausted its search traffic turned out to have a second, entirely unrelated use.
- The production bottleneck disappears. Whatever you think of the output, the distance from an idea to a finished master is now hours.
- Worldwide distribution is available to one person. No label, no studio.
- It reaches a different audience than text does. The TikTok usage is the clearest evidence of that.
And the problem that is left
It is not the music. It is everything after the music.
Removing the production constraint does not remove the attention constraint — it relocates the whole difficulty there. Nobody hears a track because it exists. Artist pages, release cadence, whatever passes for promotion, the decision about how much of a finite week to spend on any of it: none of that got easier, and all of it is now the entire job.
This is the same lesson blogging teaches, arriving from a new direction. When making things becomes cheap, the scarce resource is not creation, it is being found. Six months in, that is the part I have not solved, and I do not think faster generation is going to solve it.
The September terms change adds a second constraint I did not have during the six months described here. With releases now metered against a monthly download allowance, unlimited generation no longer implies unlimited output — you have to decide which tracks are worth spending a release slot on. Given that I could not keep up with promoting what I was already putting out, a ceiling on releases may turn out to be the less painful of the two limits.
Worth doing anyway. Twenty years of writing turning out to be the foundation for something I had given up on is not an outcome I would have predicted, and I am still going. Blog and music both, mixed together, as the same body of work.