Almost nobody sits down to write and finds their head empty. What they find is an empty project, a cursor, and no first move. Those are different problems, and only one of them is solved by the thing most people reach for.
16 August 2026. Written for people who make music at home, and for anyone who has stared at bar one for an hour.
Ask a producer what is wrong and they will say they have run out of ideas. Watch them work and something else is going on. They know roughly what they want. They can hum it. Three records get named as the neighbourhood it should sit in. The thing they cannot do is play the first eight bars, because the first eight bars have to be decided from nothing, and every decision rules out a hundred others.
That is not a creativity problem. It is a starting problem, and it hits beginners hardest for an obvious reason: an experienced producer has a stock of first moves they have used before, so a blank project is a familiar room. If you have made twenty tracks, you have twenty ways in. If you have made two, you have two.
It also explains the most common beginner pattern there is, which is fifty projects started and none finished. Each one dies at the same place, a little after the start, when the initial move runs out of road and there is nothing behind it.
There is now an obvious escape. Describe the track, wait, and get a finished one. It works, in that something comes out. And when you have been staring at bar one since nine, it is very hard to argue with.
Producers have noticed. A survey of 1,200 music creators this year found 87% had used AI somewhere in their process, from writing to promotion. Not a fringe, then. Nearly everybody.
The interesting part is what the same people say they are worried about. More than a third said leaning on these tools risked compromising what they actually meant to make, or thinning out what made their work theirs. Sound On Sound's own reporting this year found producers describing bread-and-butter skills, balancing, corrective EQ, dynamics, drifting into optional, because something else picks them up.
So the picture is not people who love it or people who hate it. It is a lot of people using it and quietly wondering what it is doing to them. Most of them do not want the machine to do the creative job. They want it to save time and stay under their hand.
Here is the trade, plainly. A prompt does not hand you a start. It hands you an ending. You skipped the part where the track becomes yours, and what arrives is finished, so there is nothing left to decide. You are not unstuck. You have just stopped.
None of this is new advice, and it is not ours. It is what producers who finish things do.
Make the first decision arbitrary. Pick a tempo and a key by rule, not by taste. 120 and A minor. It does not matter, because the point is to be past the decision, not to be right. Almost everything you commit to in the first ten minutes gets replaced anyway.
Write the sentence before the music. One line describing what this is: a slow, close, sad thing at 88 with brushed drums. That sentence will do more work than an hour of scrolling presets, because from then on every choice is either in it or out of it.
Three parts, then stop adding. Drums, bass, one melodic thing. Finish an arrangement with those before you allow yourself a fourth. Most stalled projects are not thin, they are crowded.
Separate the jobs. Writing and mixing are different work using different parts of your head, and doing them at once is the single most reliable way to spend four hours and finish nothing. Write badly first. Fix it later.
Finish something bad on purpose. The skill you are missing is not writing, it is finishing, and it only comes from finishing. Ten completed poor tracks will teach you more than fifty promising eight-bar loops.
Steal a structure, not a sound. Take a record you like, write down where its sections change, and build something completely different to that map. Nobody can hear a borrowed arrangement.
There is a real difference between something that gives you a first move and something that gives you a finished track, even though both are sold as help.
A tool that hands you a starting point leaves the work of judgement with you. You still choose what to keep, what to throw out, what it becomes. You are still the author, and at the end you know why every part is there. A tool that hands you a finished piece has already made those choices, and the only decision left is whether to accept it.
This is the line we build on. Our instruments write, but they write a beginning, not an ending. ARGISH composes an evolving piece from a chord wheel and a mood, and you shape it. REHEAT writes a bassline and gives you the MIDI so your own bass plays it. SILT runs loops through a tape machine until they turn into something you did not plan. None of them use AI or trained models. They use music theory and DSP, which is a duller sentence and a more honest one.
You can play all of them free in a browser, with no download and no account, on the demo pages. Whether that helps you or not, the advice above stands on its own, and it is free.
Since the whole argument here is that you are short of starts and not of ideas, it would be poor form to end without handing you one. Press the button. You get one concrete first move for a track. No sign-up, nothing generated, nothing sent anywhere. It is a list we wrote and a shuffle, which is the honest version of this kind of help.
Press the button and start with whatever it says.
Rule of the thing: do what it says for five minutes before you judge it.
Being stuck is not evidence that you have nothing to say. It is evidence that starting is hard, which it is, and always was, for everyone. The producers who finish things are not the ones who never get stuck. They are the ones who have collected more ways to begin.
Collect a few. It is a better investment than a subscription that finishes your songs for you.
We write these when there is something worth writing down. One email when a new one lands or a new Tunary instrument ships. No newsletter, no schedule.
Tunary