The No Names ProjectAI is a tool. Here's the whole workshop, no tricks: what I bring and what the machine brings.
| Me | The machine |
|---|---|
| The idea and the theme of every song | Nothing. No idea, no song. |
| The lyrics: I write them, correct them and decide how they end up | Corrects, suggests, and sometimes guides me when all I have is the idea |
| The sound in my head, the references, the direction | Generates audio following that direction |
| The voice: the model is trained on my real voice, in several timbres and nuances | Sings with that voice |
| Listening, discarding and picking the take | Another take. And another. And another. |
| Changing parts, verses and how it should sound | Redoes what I ask |
| Track order, the covers I approve, what gets released | Researches, drafts images and checklists |
It all starts with the idea I want to get across. If one day I want to talk about eating corn flakes for breakfast, I write about corn flakes. AI has nothing to say about my corn flakes. I do.
I write the lyrics in Spanish. AI helps me correct and update them; sometimes I only have the idea and ask it for guidance, then I correct until it says what I meant. I have skills I built myself for that: to correct, to guide and to kick off generation.
I think of something close. Nirvana's "Downer", for example: I have that sound in my head before the song exists.
I play guitar to hear a chord or a progression. If I want to guide how the lyrics should sound, I hum them. Then comes direction: it is a sarcastic song, I want it to sound weird, I want to mix rap, disco and even reggaeton. Yes, reggaeton. Breathe.
It isn't a copy or a recording of me pasted on top. Suno starts from its own model and trains it on my voice. The custom model (TheNoNameWay) wasn't trained on songs I had already made: it was trained on my real voice, recorded in different timbres and nuances. That's why it ends up deeply anchored to how I sing.
On top of it go the voices I trained in Suno, each with its own character:
SilverTable: named that because I sang "Ana's Song" by Silverchair, another grunge band I love, to record it. It is the voice of El Hermoso Caos. Armand Bianco: the raspy, somewhat deep one, the "Rape Me" one. Chrnx: the ups and downs, Billie Joe Armstrong style. Oh yes: punk.
That's why it sounds so much like me: it is trained on my voice, in layers. The machine does not invent who sings; it just stays in tune better than I do.
I made the record with Suno v5.5. Today it is on v6, and 5.5 has been discontinued: that is how fast the tool moves.
It can be 4 takes, it can be 10. Then more come, like covers of the same song. And I listen with a checklist in my head:
Does it have the weirdness I want? Does the voice sound odd in falsetto? Does it sound raspy? Sarcastic? Does it express the feeling? Is it Drop D or close to it?
And even then, I will most likely make another 10. If a verse or a part does not land, I change it, or change how it should sound.
Sometimes Logic Pro on the Mac, sometimes Suno Studio, sometimes code. All 13 tracks of El Hermoso Caos went through Suno Studio 2.0, and each one was loudness-checked before upload.
"Julieta" ended up about 4 dB louder than the rest of the record. On purpose: it is the song's fury. Not everything gets flattened to please a number.
I plan a record and plan which songs to write. I use AI for brainstorming, planning and my skills. Every song lives in git, in its own worktree, like a code branch: it gets tested, reviewed and merged when ready.
Claude for research (and to build this site with me). Codex for images, covers and guides, like the DistroKid one.
Basically: cutting down the anxiety of not having a clear guide. For a brain like mine, a clear guide is not a luxury; it is the difference between finishing a record and leaving it in the to-do folder.
With every record, the direction of each song gets more obsessive. Look at the difference in how I ask the machine for a chorus.
How the chorus is requested:
[Chorus]
[Screamed]
[Raw]How the chorus is requested:
[Chorus | Explosive Mockery | Raspy Shouted Lead with Gang Replies | Distorted Guitar Wall, Groovy Bass, Open Crash Drums | Full Dance-Punk Release | Saturated Live-Room Grit]None of that is decided by the machine. I decide it, line by line.
You are so cool
Like a billboard selling empty space
From a song that is coming: "You Are So Cool".
If a person who sees the world differently makes you shake for using tools, that says more about you as a musician… or maybe about the dollar sign in your music.