Measure real progression from speech, on your trial’s own scale
Speech read from recordings your trials already hold, site-visit recordings or home sessions. The reading stays stable when phones or their software change, recalibrated from the audio itself, and every recorded rating interview is checked, so a change of phone or rater does not read as disease change. How readings stay stable across phones ›
Talk to our team
Built for regulated clinical trials
Where CNS trials lose signal
A drug effect has to show through coarse scales, rater variation, many sites, changing phones and unread recordings.
Coarse scales, few visits
Rating scales move in steps and are given at a handful of visits, so slow change shows late.
Rater variation
Baseline inflation, drift and interview differences add noise that can look like placebo response.
Recordings left unread
Trials already record rating interviews and speech tasks, and most of that audio is never reviewed or analysed.
Sites, languages, phones
Multinational trials rate the same scale across sites and languages, and over a long trial phones and their software are replaced. Each adds variation that is not the disease.
Why speech
Speech carries motor, cognitive and psychiatric signs
Speaking uses the brain, breathing, the voice box and the muscles of the mouth together, so one short session can hold all of these signs.
More readings per participant
Existing and site-visit recordings add a reading at each visit; home sessions add readings between visits. In dose-finding, each arm gets more readings before the dose decision, unblinded view only.
Nothing to wear
Speech comes through your eCOA or app vendor, a phone call, or our own capture where none exists. No wearable to ship or charge, and no extra site visit.
State of the art, and no black box
Named voice and language features behind every reading. The reference model is calibrated on your screening and baseline ratings, blind to arm, and locked before post‑baseline recordings are read. For your statistician ›
Evidence
White papers and posters
Our methods are presented at CTAD, ISCTM and AD/PD. White papers on Alzheimer’s, Parkinson’s, Huntington’s and ALS are available on request.
Speech‑based assessment in Alzheimer’s disease and MCI
2026 · Request White paperSpeech‑based assessment in Parkinson’s disease
2026 · Request PosterNoise‑cancellation variability as a source of longitudinal drift in speech‑based digital endpoints for Alzheimer’s disease trials
CTAD · Boston · November 2026Speech readings on your scale
Coarse scales and few visits hide a drug effect. You get a reading on your own scale at every recorded visit, joined to the visit score and fixed in your SAP, in units your reviewers already know.
See speech biomarkersEvery recorded interview checked
Baseline inflation and rater drift add noise that reads as placebo response. You get flags per interview, rater and site from screening onward, so your blinded team can act before randomization. Cephalgo never changes a score.
See Trial QCWays to start
Pick a starting scope
Start from recordings you already hold or from baseline in a new study. You know the fee and timeline in writing before any data moves.
Re‑analysis of a trial you already ran
Lowest-risk startYou get readings on your scale, rater flags and variance components from audio you already hold, and from a completed trial an effect size your statistician computes on the locked readings. No new collection.
Prospective study
Trial QC
Pre-screening
New to speech? See what speech adds, from recordings you already hold ›
Already using a speech vendor? Get a second reading on the same audio, on your scale; your vendor stays ›
Running your own speech models? Get named features to test your own models against ›
