Clarity · Cognitive voice biomarkers
CaringAI Listen already scores the answers.
Clarity reads the voice that produced them.
You already pull cognitive markers out of speech on every screen: fluency, coherence, response latency. Clarity reads the layer under that, the acoustic and timing signature of how the voice assembles the words at all. Same phone screen, same MyCog and PHQ-9 your team already administers, one more read on the way out.




Rides along with the instruments already in the call
Four things in the audio a score cannot hold.
The hesitation that lands just before a noun, the filled pause, the swap to a vaguer word. A patient who recalls two of three words scores the same whether it took her one second or five. The seconds are where the change shows up first.
Turn latency, pause distribution, and how the rhythm of a reply degrades over a nine minute call. Fatigue across a single screen reads differently from a stable baseline, and it reads long before a total score crosses a threshold.
The pitch and intensity variation that flattens with apathy and low mood. Sitting next to a PHQ-9 that the patient self reports, it is an independent read on whether the answers and the affect are telling the same story.
Consonant sharpness and vowel stability, the motor speech layer that separates a vascular picture from a neurodegenerative one. Not a diagnosis. A reason to look, on a call you were running anyway.
Six clinical signals from one audio sample, powered by Amplifier Health's Sona-2 model
Illustrative voices from a Clarity pilot
“Her total was the same number three screens in a row. The latency trend was the only thing telling us she was working much harder to get to it.”
“I have eleven minutes with these patients. Anything that tells me which of them to spend the eleven minutes on is worth more than another form.”
“We were already recording every call. Finding out there was a second read sitting in the same file, and we just werent listening for it, was the part that got me.”




Privacy
Your voice is yours.
Screening audio is handled the way CaringAI already handles it. The call is recorded and processed for the read, the signal is written next to the score, and nothing leaves the pipeline your team already runs. On the demo page below, the audio never leaves the browser at all.
One minute from now.
- 01The screen happens exactly as it does today
Outbound call or patient hotline, five to ten minutes, validated instruments delivered by your voice agent with the consistency you already sell.
- 02Clarity reads the same recording
No second call, no extra questions, no change to the patient experience. Sona-2 takes the audio your pipeline already produced and returns a cognitive voice read on it.
- 03The read lands beside the score
One more panel in the screening record, one more line in the care plan, and a trend across screens rather than a single number at a single point in time.
Early access
Clarity is ready to pilot inside CaringAI Listen.
A cognitive voice read on every screen your platform already runs, from audio you already capture, built on Amplifier Health's Sona-2.