Skip to content
Demo · simulated output

One recording, three ways of listening.

One recording of speech, read by three kinds of model. Press play and watch what each one is able to take from the voice while it runs — or upload one of your own samples and run it through the same board.

Signal breadth.

One recording · three systems listening

00 / 42
Sona-2 markers active
Inbound call · 43 s0:00 / 0:43

Routine check-in call, consumer handset, single speaker

01 / Transcript00 / 42

Words only · nothing about the speaker

02 / Emotion API00 / 42

Affect labels · no baseline · no evidence tier

03 / Sona-200 / 42

Six biological systems · measured against her own baseline

Sona-2 systemsRespiratoryPhonationAutonomicProsody & timingCognitiveAffective

The read assembles as the recording plays

Source

Illustrative demo. Outputs shown are simulated for explanation and were not produced by calling any product live. Amplifier is not FDA cleared, does not diagnose, and returns wellness and context signals only. Uploaded audio is decoded in this browser only.

What separates them

A label is not a measurement.

Baseline
Affect labels are scored in the abstract. Sona-2 scores against the speaker's own history, which is why “below her range” means something the first list cannot say.
Evidence tier
Every signal arrives with the strength of the science behind it — established, emerging, investigational — so a care team knows what to act on and what to note.
Continuity
The read is retained and comparable across calls. Voice becomes a longitudinal measure rather than a per-turn impression.
Next

Run this against your own population.

A paid discovery engagement establishes which signals hold in your recordings, at your audio quality, for your population — before anything is built.