How we select an AI track on inHits
Behind the scenes · June 2026
Making music with artificial intelligence is easy these days. Making music that’s actually worth listening to is a different story. That’s where our work begins.
Every day, thousands of AI-generated tracks are born. Most of them will never reach anyone’s ears — and that’s how it should be. The problem isn’t quantity: it’s the lack of a filter. inHits exists to be that filter. In this article we tell you, plainly, how a track goes from a generated file to a hit you find in the app.
1. The proposal: from our software
It all starts with our proprietary software. We don’t just “press a button”: the system works on the whole track — the lyrics, the style, the genre, the structure — aiming to propose high-quality new music. It’s a starting point, not a finish line.
At this stage there are no “hits” yet. There are candidates: many tracks, many possible directions. None of them ends up in the app automatically.
2. The listening: human ears, not metrics
This is where the part that sets us apart comes in. Every candidate is listened to by real people. We don’t look at a machine-generated score: we listen. A song works or it doesn’t in the same way it does for you when you hear it for the first time — either something grabs you, or it doesn’t.
Our curators are industry professionals. Among them are people who have scored soundtracks for documentaries and series, who have won international awards, who have spent years understanding why some songs stick and others don’t. They’re the ones who decide.
3. The decision: publish, improve or discard
After the listening, each track takes one of these paths:
- Publish: the track is ready. It’s a hit, it enters the catalog.
- Edit or improve: there’s potential, but something needs fixing. We step in and listen again.
- Discard: not for us. The track doesn’t pass the selection and will never see the app.
Most candidates get discarded. It’s not a flaw in the process: it is the process. If we published everything, we’d be exactly one of those scattered, noisy places we want to stand apart from.
What makes a track a “hit”?
There’s no formula, and we’re wary of anyone who says otherwise. But there are questions we always ask: does the song have an identity? Does it stick in your head? Would you play it again? Would you recommend it to someone? If the answers are yes, we’re probably looking at something worthwhile.
Why a human must have the last word
Music speaks to a part of us that no metric can truly measure. AI is an extraordinary tool for proposing; but deciding what deserves to be heard remains a human choice. That’s why, on inHits, the final yes is never given by a machine.
It’s a long job and, yes, we discard a huge amount. But when you press play on inHits, you know someone listened to it before you and thought: this one, yes.