A little voice check.
A lot of gentle science.
HAPPO listens to what you say and how you say it — then offers a small, kind plan you can do in five minutes. Here’s exactly what happens behind the scenes — and what HAPPO will never do.
Speak in your own words.
Speak in any of nine languages. There’s no script and no questionnaire. Just say what’s on your mind, in your own words.
A tiny heart check.
HAPPO listens to small clues in your voice — pitch, energy, rhythm, waveform patterns — alongside what you say.
Eight emotions, weighed.
Your message is mapped against the Plutchik wheel — joy, sadness, trust, fear, surprise, anticipation, anger, disgust.
A gentle plan, just for you.
Pick a five-minute practice — breathe, listen to instrumental music, reroute a thought, notice three good things, or simply keep talking.
Words tell us a lot. Voice tells us a different lot. HAPPO measures small acoustic signals — none of them is a diagnosis. They’re context that lets her respond more kindly.
range.
The narrower your voice’s pitch swings, the more we listen for low energy. The wider the swing, the more likely your voice is animated — for better or worse. HAPPO never tells you what either means; she uses it to choose a softer or more energetic response.
energy.
How forcefully you push air through your vocal cords. Higher energy can mean excitement, frustration, or rush — context matters. Lower energy often signals tiredness or sadness. HAPPO uses energy to pace herself: more silent listening when you’re low, more engaged when you’re activated.
rhythm.
How regularly you pause, how long your pauses last, and how steady your speech tempo is. Disrupted rhythm — sudden pauses, rushed bursts — can be a small signal that something’s on your mind. HAPPO slows her own tempo to match yours.
Robert Plutchik proposed eight primary emotions in 1980. HAPPO uses a modern variant of the same wheel to understand the emotional shape of what you say — not to label you, but to choose her words.
- Joy + anticipation · 41% of last week’s conversations
- Trust · 23% — usually before you open up
- Fear, surprise · 14% — HAPPO slows down here
- Sadness · 13% — practices like 4-7-8 and Three Good Things show up here
- Anger, disgust · 9% — HAPPO listens longer before offering anything
These are contextual signals, not diagnoses. HAPPO never tells you you have anxiety, depression, or any clinical condition. If you want a real assessment, please see a real clinician — she’ll happily help you find one.
We didn’t reinvent the foundations — we stood on them and chose carefully. Here’s what powers HAPPO under the hood.
Whisper for transcription, HAPPO.ai for understanding.
Voice to text via Whisper (open-source, operated for us by Replicate; we send the recording and nothing else, and it expires on their side within about an hour). The acoustic and language analysis both run on our own self-hosted models. Conversations are generated with the help of established AI model providers, encrypted in transit and at rest. We do not keep the raw audio once the features are extracted. The pipeline builds on our peer-reviewed research (doi.org/10.1007/s10015-023-00852-4) and has improved since.
Custom voice biomarker stack.
Pitch, energy, rhythm and pause extraction, refined on HAPPO.ai plus a small in-house classifier trained on 230,000 anonymised samples. This classifier runs on our own servers, and the signals it produces about you never leave them.
Encrypted on your terms.
AES-256 at rest, TLS 1.3 in transit. Hosted on GKE in Singapore, with EU mirror in Q3 2027. Auditable by request.
All that, in your pocket.
No diagnosis. No verdicts. Just a kind hippo who listens, in nine languages, any hour.