An estimate of how a typical brain responds to this music, based on brain-imaging research. Not a scan of you, and not medical.
What you're seeing
A 3D model of a human brain from an open anatomy atlas. While a song plays, areas light up to show how strongly a typical listener's brain is estimated to respond at that moment. Warmer and brighter means a stronger response.
Below the brain: the sound itself, as pitch over time, and the response of eight brain areas over the last few seconds. Click any area to read what it does.
How the music becomes brain activity
Each song is analysed 20 times a second into 128 frequency bands, from deep bass (30 Hz) to high treble (8 kHz), plus loudness, beats and sudden changes.
These sound features are mapped onto brain areas that studies link to hearing and rhythm: the hearing relay in the brainstem and thalamus, the hearing cortex, nearby sound-pattern areas, and the movement and timing areas that follow the beat. Only these groups can light up. Within them, each part follows what it is known for: the left hearing cortex leans towards fast timing such as drum hits, the right towards pitch and tone, and the primary hearing area, tucked inside the fold on the side of the brain, is laid out by pitch, so you can see low and high notes land in different places. The mapping draws on these studies:
What's real and what isn't
The brain anatomy is real atlas data, and the sound analysis is computed from each song. The brain response is an estimate: a research-based mapping driven by the sound. It is not a measurement, it describes a typical brain rather than any one person, and it is not your brain.
We also trained a small model on real brain scans of one person listening to music (the open dataset OpenNeuro ds003720). It did not beat its baseline tests yet, so we call it experimental. It runs in your browser for pasted links and your own files; the built-in playlist uses the sound analysis only.
Timing
Brain scans (fMRI) measure blood flow, which lags behind the music by about 4–6 seconds. By default we show the fast "neural" timing, lined up with what you hear. The experimental model can also be shown with a 5-second blood-flow delay, under "Use your own audio / advanced".
Your privacy
Audio files from your device are analysed in your browser and never uploaded. A pasted YouTube link that isn't in the playlist is analysed on our server: the audio is fetched briefly, turned into sound features, then deleted. No microphone, screen or tab access is requested unless you choose "Listen to another tab".