Baseline calibration
A relaxed 15 minute session, at your own pace, so VocaLoci gets to know your voice. There is no test to pass and nothing to get wrong.
Real-time speech support
You already know what you want to say.
VocaLoci listens while you speak and offers the word you were reaching for, usually within half a second. It supports your voice. It never speaks for you.
See how it feels to use.
A short recording of a real session, so you can see exactly what happens and decide for yourself.
Recorded session, 58 seconds
Most speech technology was not built with you in mind.
Mainstream tools tidy away the moments that matter most to you. We think those moments deserve support while they are happening, rather than a correction afterwards.
- What you deal with today
- Everyday speech tools smooth over disfluency, so what you actually said gets lost.
- One model for everybody, when no two people speak quite the same way.
- Help tends to arrive after the moment has passed, when it is no longer useful.
- What we do differently
- Support that arrives in the moment, while it can still help.
- A model that learns your voice, and keeps learning it.
- Everything runs on your own device, so your voice stays private.
No two voices are alike. Neither are our models.
From your very first session, VocaLoci is learning how you speak rather than applying someone else’s average.
Adaptive learning
With your consent it keeps learning from how you actually speak, so it fits you a little better every time you use it.
Proactive assistance
It learns which situations tend to feel harder, like calls or presentations, and is ready before you get there.
More than just listening.
Two of these work today. Two more are being researched, and we will only ship them if they genuinely help.
-
Audio and acoustics
VocaLoci listens to the sound and rhythm of your speech through your microphone.
Live -
Language context
It follows the sense of your sentence, so any suggestion fits what you were actually saying.
Live -
Facial cues
An optional camera input could notice a hard moment forming. Entirely your choice, and off unless you turn it on.
In research -
Wearable signals
Signals from a smartwatch, such as breathing, may give a few seconds of warning before a hard moment.
In research
Early pilot findings
What early users tell us.
Early days, and a small group so far, but this is what people have reported back to us.
40%Less mental effort while speaking
35%Less worry in the run-up to speaking
2xMore of the conversations they wanted to have
On these numbers. These are early pilot findings from a small cohort, not a controlled trial, and we report them as directional rather than definitive. The engineering figures are firmer: prediction latency of 491 to 508 milliseconds measured in live sessions, and top-3 accuracy scored on a frozen 181 item evaluation set. We would rather show you the method than the marketing, so come and find us at the conference and ask.
So many people, and so little built for them.
- 1.6B+People worldwide live with a communication difference
- 400M+Live with a speech difference
- 80M+Stutter
- $186BEstimated annual loss to the US economy from communication disorders
- 67%Unemployment rate among people with severe speech differences
Built inside the Cambridge research ecosystem.
Early access
Join the waitlist.
We are welcoming people in small groups, so we can set VocaLoci up properly with each person. Tell us where to reach you.
- An early place when the next group opens
- A friendly setup session, at your own pace
- A direct line to the people building it
- No spam. We write only when there is something worth saying
You’re on the list.
Thank you for trusting us with this. We will be in touch when the next group opens. If you are at the conference, please come and say hello.