The app that turns a hummed tune into sheet music is finding an unlikely home in India

A California startup built to lower the barrier to music education is resonating in ways its founders did not fully anticipate.

I recently sat down with Andrew Carlins, an MBA classmate of mine, who told me about the company he is building, Songscription. Songscription describes itself as an AI lab for music learning. Its flagship product allows users to upload an audio recording and convert it into sheet music, MIDI, tabs, and piano roll. Users can also tailor the output (e.g., sheet music) to their level of play. You upload a file. You get notation back. That is the whole product, and it turns out that that is enough.

Carlins grew up with a stutter and realized that when he sang, his stutter disappeared. “Music is quite literally how I found my voice,” Carlins explains. He built Songscription because he believed that a world with more access to music learning and live performance is a better world, as he had personally experienced the unique ability music has to empower individuals and unify people from different backgrounds.

Songscription’s users tell the story even better than Carlins or the product description does. A music teacher in Brazil is using Songscription with elementary students as a teaching supplement. A violin teacher in the United States used it to produce notation for a student who learns better from the page than by ear. A visually impaired pianist used it to get his own compositions written down for the first time. One feedback message a user wrote to the Songscription team said simply: "I never thought this would be notated. Thanks for fulfilling a dream." None of these people needed a machine to create music for them. They needed one that could help them find their way to learn and perform music live.

The India story is worth telling in some detail. Sushma is a Carnatic music teacher whose students, like a lot of students in India right now, listen to as much film music as they do classical. They come to her lessons with requests for Tamil and Hindi film songs, melodies they already know by heart and want to be able to read on the page. The problem is that Carnatic music and western staff notation are two distinct systems, and Sushma, trained in one, had no particular grounding in the other.

She found Songscription, uploaded a film song her students had been asking about, and got readable sheet music back. She wrote to the Songscription team afterward, not as a formal review but as a note. She mentioned the output was clean, that the process had been easier than she expected, and that the messages on the website while her file was processing had made her laugh. She closed with regards to the whole team. It was a small note, but it is the kind of feedback that tells you something about whether a product is landing and having the impact it was built to have.

What Sushma's experience points to is a gap in how music technology has been built. India has a large and serious music education culture, a tradition of both classical training and mass popular music, and a growing number of learners who move between those worlds. Most transcription tools were designed with western music pedagogy as the default. A Carnatic teacher navigating film music notation is not a use case those tools were built for. Songscription, which gives users access to western notation regardless of their musical background or tradition, turns out to fit that gap reasonably well. Carlins says the company eventually wants to support non-western notation systems as well, though that work has not shipped yet.

It is worth being clear about what Songscription is not, because the AI music conversation right now is mostly about generation. Products that compose new music, write lyrics, or produce tracks from a text prompt are getting most of the attention. Songscription does none of that. It takes music that already exists and makes it readable and accessible. That distinction matters for how you evaluate its effects on musicians. Songscription is not trying to do the creative work for anyone; it is trying to remove the technical barrier between a musician who has an idea and a page that can hold it.

The accessibility theme runs through the user base consistently enough that it seems like more than a coincidence. The visually impaired pianist who could not use conventional notation software. The Carnatic teacher whose training did not include western staff notation. The elementary school students in Brazil whose teacher needed a faster way to produce arrangements. Each of these is a person who, for a specific and legitimate reason, could not get what they needed from the tools that already existed. Songscription is not solving every version of that problem, but it is solving enough versions of it that the pattern is noticeable.

The product currently works best with single-instrument recordings, with piano as its strongest use case and other instruments available in beta. Full band transcription is not yet supported. Songscription is publicly available, with a free tier, at songscription.ai.