A
LoadingThe technology
Everything we know about the technology behind your report: where it comes from, what it measures, what we have tested ourselves, and where its limits are. Nothing is hidden, and nothing is claimed that we cannot show.
Why a voice
Speaking is a physical act. Your breath, the tension of dozens of small muscles and the way your nervous system regulates them all shape the sound, before you choose a single word.
Emotion, stress, tiredness and mental effort change those processes, and the changes can be measured in a recording. Voice research has studied this for decades: pitch, the steadiness of the tone, pace and pauses all move in recognisable ways.
AVOCO applies that research to two questions: what kind of person the voice belongs to, and what state they are in right now.
What is measured
From the recording, the analysis extracts acoustic measures of the same kinds used in speech science:
01
The fundamental frequency of your voice, and how much it varies.
02
Tiny cycle-to-cycle wobbles in pitch (jitter) and in loudness (shimmer).
03
How clear the tone is against breath and noise: the harmonics-to-noise ratio.
04
The shape of the sound's spectrum: MFCC, spectral centroid and spectral entropy.
05
Your speaking tempo, and the length and frequency of your pauses.
06
The rise and fall of intonation across phrases, and how loudness is spent.
No word is transcribed, stored or understood. What you talk about is entirely up to you.
From recording to report
Nothing is analysed until you tick the consent box. The recording is prepared on your own device in the format the analysis expects; from a video only the sound is used, and the video never leaves your device.
The analysis evens out the level, reduces noise and cuts silence, so that it measures your voice rather than the room.
The acoustic features described above are extracted from what remains.
An ensemble of machine-learning models, neural networks working alongside classical algorithms, turns the measurements into a score from 0 to 100 on each of eight personality types and fourteen emotional scales.
We present the scores with AVOCO's official description of each type, and add readings of our own, such as your best-fit fields of work, each marked as ours.
Where it comes from
The analysis is developed and run by Voxera, a voice-AI company in Kazakhstan. Its creator, Vsevolod Shchegelsky, has described the method in a published interview; the cleaning, the acoustic features and the model ensemble on this page come from that description.
The eight personality types and their descriptions are AVOCO's official typology, from its Vocal Psychotyping System. We publish them as they were written.
AVOCO USA runs this platform: the recording, your account, the protection of your data and the report you read.
The first two figures are reported by the developer; the time is what we measure.
How far to trust it
A trustworthy tool says how it was tested. Here is everything we know, including what has not been published yet.
In the developer's studies
These are the developer's own figures. No independent study has been published yet, so we present them as reported, not as proven.
In our own tests
✓ Repeatable
The same recording returned exactly the same numbers, to the decimal, every time we sent it.
✓ Stable where it should be
Speaking faster or slower, louder or quieter, or saying something completely different did not change the personality result. It follows the voice itself, which is why the same person gets the same type on different days.
✓ Responsive where it should be
The emotional scales move with pace, pauses and phrasing, as they are meant to. They also react to the microphone, background noise and the room, which is why we help you record well.
✓ Honest about a single recording
A voice close to the border between two types can land on either. That is why we average all your recordings and show how settled your type is.
What it is, and what it is not
Your voice stays yours
For the truest reading
Sources
Thirty seconds, on any topic. The full method stays with you, at the end of every report.