About
Every note traces
to a measurement.
Iris.Art turns the body you already have — your iris, your voice, your pulse, your face, your handwriting, your prints — into a piece of music. Not a random one. One where every pitch, every rhythm, every instrument choice has a documented reason, shown to you in plain language on the same page as the composition.
The problem
Generative music is usually a black box.
Feed most music-generation tools a prompt or a photo and you get a plausible-sounding result with no way to ask why it chose what it chose. That's fine for a lot of use cases. It's not fine for something built from your own body. If a piece of music claims to be “made from your iris,” you should be able to see exactly how — not just trust the claim.
The approach
A two-layer model: one interpretable, one not — and we're honest about which is which.
Every capture runs through classical, documented computer-vision measurements — the crypt pattern in your iris, the pitch contour of your voice, the stroke geometry of your handwriting. These measurements, and the equations that turn them into specific musical decisions, are the thing we show you in the derivation table on every composition page. Same capture, same model version, same music, every time.
We also run several modalities through modern foundation models — DINOv2, ArcFace, WavLM, TrOCR — to get a richer numeric fingerprint of your capture. That signal is real and useful, but it's opaque by nature, so we never let it drive a musical decision directly. It sits alongside the composition as a comparison and exploration tool, always labeled as what it is.
The shape of the product
Nine modalities, and it's still growing.
Iris, voice, pulse, face, handwriting, fingerprint, palmprint, periocular, and rPPG-face. Two of those pairs — iris with periocular, face with rPPG — come from a single capture and produce two compositions instead of one. Fingerprint reads all five fingertips and blends them democratically rather than picking a favorite.
Where it runs
Web, iOS, and Android, kept in sync.
Start a capture on your phone, see it finish on your laptop. Your dashboard updates live as each pipeline stage completes, because a biometric capture experience shouldn't feel like refreshing a status page.
What we take seriously
Biometric data is not casual data.
Every raw capture is encrypted at rest before it's stored, with a key that never reaches your browser or your phone. Every upload requires explicit, versioned consent. Illinois residents are blocked from capture flows entirely, rather than us trying to thread a compliance needle we're not confident in. If you connect a DNA profile from our sibling product, LifeSong, that's a second, separate opt-in — off by default. The specifics are in our Privacy Policy, not buried in a FAQ.
See it work
Read the general shape of the mapping.
We don't publish the exact equations or version constants — those are the thing we build on — but the general model is public.
The Science →