About the project
An open tool, and an open commons, for the raagas
twelveswaras names the raaga in a piece of Carnatic or Hindustani music, and turns what it hears into open data that makes it better for everyone. It is free, and it belongs to no company.
What this is
Play it about a minute of Carnatic or Hindustani music and it names the raaga, tells you how sure it is, and shows you how to hear that raaga yourself. It is a non-commercial public good: no company owns it, there is no login, no ads, and nothing about you is sold.
The twelve swaras are the twelve note-positions of the octave, the shared alphabet of Carnatic and Hindustani music. That is the name, and the intent: both traditions, one tool. Carnatic came first; Hindustani is now in preview (its reference facts are checked against published sources; not yet reviewed by a musician).
Who is behind it
twelveswaras is built and maintained by Sathya Sankaran, who builds open commons: public goods that belong to everyone who uses them. The raaga system is shared inheritance, carried forward by generations of musicians and teachers, and the tools for hearing it should be shared the same way. The model, the code and the reference data are all open, free to use, check or build on.
It is also built to be added to. A correction when it hears a raaga wrong, a recording of one it does not know yet, a clearer explanation on a raaga page: each one makes the next person's answer better. That is what a commons is for, and everyone is welcome in it.
What happens to your recording
This is the part that matters most, so here it is plainly.
When you identify a raaga
Your audio is analysed to find the melody and the tonic, then discarded. It is never stored. The only thing kept is the anonymous result, which raaga and how confident, never the sound.
When you choose to contribute a clip
You are donating it to make the recogniser better. The clip is stored privately and used to check the label and train improved models. We publish the model and the features we learn from it, not your recording.
If you release a clip publicly
Only when you tick "release under CC-BY", and not otherwise, does your clip join a public, openly-licensed dataset, with credit to you. That choice is always yours.
You only contribute clips that are your own performance, or that you have the right to share. You decide whether the audio is ever made public. To have a contributed clip removed, email us and it will be removed.
How it works
There is nothing to set up beyond the audio. It follows the main melody, finds the tonic (Sa) from the drone, and builds a picture of how the melody moves between notes, the gamaka, not just which notes are sung. A model then names the raaga with an honest confidence, so a shown "70 percent" is right about 70 percent of the time. It needs a drone: a concert or a broadcast works well, a solo voice with no tanpura is unreliable. The full method, with citations, is on GitHub.
What it can and cannot do yet
Seventy raagas: forty Carnatic and thirty Hindustani, a mix of the melakarta, janya and thaat raagas you hear most often in concerts. The reference facts on the raaga pages are checked against published sources; not yet reviewed by a musician. It is accurate on clean studio recordings and less so in the wild, and closing that gap is exactly what the commons is for: every clip people contribute helps. The "how to hear it" notes on each raaga page are a first draft, and we are looking for Carnatic and Hindustani musicians and teachers to help refine them. We would rather tell you this than overclaim.
How the licensing works
The code is open source under the MIT licence; the running recognizer is AGPL-3.0, because it uses the Essentia audio library, so anyone who runs a modified copy for others must share its source. The model weights are CC-BY-4.0: the model learned only from the openly licensed Raga Recognition Dataset, and the weights are downloadable from GitHub. The community data commons is CC-BY-4.0 too, kept separate from non-commercial research recordings so it stays cleanly reusable by anyone. Cite twelveswaras with doi:10.5281/zenodo.24096813. Full detail is in the Governance and Methodology on GitHub.
How to help
- Contribute a clip after you identify a raaga, or if you already know it, contribute a recording directly. It is the single most useful thing, and it makes the tool better for everyone.
- Review the raaga reference if you teach or perform, so the learning notes become trustworthy.
- Contribute code or ideas on GitHub, where the whole project is open.
With thanks
Shaale contributed the first recordings to the commons, student performances shared so the recogniser can learn from them. They cover raagas it cannot name yet, which is how its vocabulary grows.
Everyone who contributes is credited in CONTRIBUTORS.md, whether their recording is released publicly or kept private and used only for training. Contributing is anonymous unless you ask to be named.
Questions
Is it really free, with no account?
Yes. Non-commercial, open-source code, an open CC-BY data commons, and no sign-up. It is a public good, not a product.
Which raagas does it know?
Seventy raagas, forty Carnatic and thirty Hindustani, a mix of the raagas you hear most often in concerts. The reference facts on the raaga pages are checked against published sources; not yet reviewed by a musician.
What happens to my recording?
If you are only identifying, it is analysed and discarded, never stored. A clip is kept only if you choose to contribute it, and made public only if you also choose to release it. The section above has the full detail.
Can I reuse the data or the model?
The code is MIT, and the model weights and the community data commons are CC-BY-4.0, all freely reusable with attribution. The running recognizer is AGPL-3.0 because it uses Essentia. The weights are on the v1.0.0 release on GitHub; cite twelveswaras with doi:10.5281/zenodo.24096813.
Can I use it from an AI assistant or the command line?
Yes: add https://twelveswaras.com/mcp as a remote MCP connector in Claude, ChatGPT, Cursor or any MCP client, with no sign-in (setup). For a recording on your own machine, the twelveswaras npm package runs npx twelveswaras identify recording.m4a, and with no arguments it is a local MCP server.