A short face scan and a one-minute voice task, read by an engine that writes a reflection in plain words. No long questionnaire.
Numa is an engine that listens to how you look and sound right now, and tells you what it noticed, in language a friend would use.
Modern models do the reading: on-device face tracking, speech recognition, and a language model that writes, never a questionnaire.
The subject is you, not a category. Every reflection is written from your own words and signals, and rated by you afterwards.
Face and voice together: head movement, expression, blink, pitch, pace, pauses, energy, and what you actually said.
Signals become five vectors and a narrative: a signature, your strengths, one blind spot, and how today differs from last time.
Centre your face in the ring. Turn slightly left, then right, come back. Tracking runs on your device; the scan only counts if it saw you do it.
Five short lines to repeat, then one question on its own screen. Forty-five seconds, in your own words. The recording is transcribed and deleted.
Face summaries, some seventy acoustic parameters and what you said become five vectors and a written reflection. Come back and it compares you only with you.
Most tools ask you to rate yourself from one to five, forty times, and get the answer you think you should give. Numa asks one question and listens to how you answer it.
Every parent knows the answer before asking. Children give the reply that ends the conversation fastest, and the ones who most need to be heard are often the best at "fine": the bullied, the anxious, the quietly sad. Numa is built for the gap between what is said and how it is said.
Done together, with a parent. The child sees everything the parent sees.
Every step waits for you. Nothing is recorded unless you actually did it.
Look ahead, turn slightly left, then right, come back to centre. Face tracking runs on your device and the scan only counts if it saw you do it. Frames never leave the phone.
Five short lines to repeat, then one question on its own screen: What will you do tomorrow? Forty-five seconds, in your own words.
Your recording is transcribed and reduced to some seventy acoustic parameters, then deleted. Face summaries, voice summaries and your words are fused into five vectors and a written reflection.
Come back tomorrow, next week, next month. Numa compares each session with your last one, so the story is about change, not a verdict.
Every session is measured close to two hundred ways, then distilled into five vectors you can actually read. Each vector has a trait reading, by design, and a state reading, right now. Scores stay in a moderate band on purpose: one session is never an extreme verdict.
Fifty-two expression channels, each tracked for level and variability, plus head pose, blink rate, stillness, symmetry and how you followed the prompts. All on your device.
Pitch and its movement, energy and its dynamics, steadiness, pace, pauses, spectral balance, formants and cepstral texture, from the recording itself.
What you actually said to one question, transcribed and read as the most personal signal of all.
Every reading compared with your own earlier sessions, so the story is about how you change, not where you rank.
How easily things move through you today: engagement, rhythm, momentum.
Steadiness of attention: how settled you look and how cleanly you speak.
Your reach toward the world: expressive range, warmth, the will to be met.
Energy and recovery: what is in the tank, and how fast it refills.
How much of the inside shows on the outside, in face and in voice.
Five is the default set. Because every vector is a reading of the same signals, more can be defined for a setting: a school, a clinic, a team.