How SuTalk Works

The 5 Dimensions of Time

Most reminder apps treat all time equally. SuTalk does not. Every event is classified into one of five cognitive dimensions — each with its own communication style, emotional register, and voice tone:

1. Daily — Habit-scale. Procedural, matter-of-fact. "Take your meds" belongs here. High frequency, low ceremony.
2. Weekly — Rhythm-scale. Supportive, anticipatory. Team meetings, exercise routines — events that anchor the week.
3. Monthly — Commitment-scale. Deliberate, grounding. Bills, reports, recurring obligations that shape the month.
4. Yearly — Milestone-scale. Reflective, emotionally warm. Birthdays, anniversaries, renewals — events tied to identity.
5. Long-term — Vision-scale. Motivational. Goals, life events, reminders set years in advance. SuTalk returns these with the original Commander intent intact.

The same words spoken in a daily tone and a yearly tone carry entirely different weight. SuTalk does not choose one voice for everything.

The Two-Voice Model

SuTalk uses two distinct personas operating in sequence:

• The Commander — the user's structured input. When creating a reminder, the Commander provides not just a time, but context: what the event is, why it matters, what the user hoped to feel when the moment arrived. This intent is stored on-device.
• The AI Companion — SuTalk's adaptive voice. When the reminder fires, the Companion returns the stored Commander intent alongside the alert — warm, calibrated to the event's dimension. It creates a dialogue, not a notification. The Companion requires connectivity; the Commander's stored intent does not.

To be precise: "Two Voices" refers to two interaction roles — the user's input role (Commander) and SuTalk's output role (Companion) — not two separate audio tracks. The Commander is the user's structured intent at creation time; the Companion is SuTalk's voice at trigger time.

Multilingual Voice — Speak and Hear in Your Language

SuTalk is designed for global users. Voice output is available in English, Thai, and additional supported languages. Users speak commands in their own language and receive spoken reminders in the same language.

This is not a translation layer on top of an English-first system. The emotional tone and communication style of each dimension (matter-of-fact for Daily, reflective for Yearly) is calibrated per language — not translated after the fact.

Offline-First Architecture

All critical alerts are stored locally and fire regardless of network state. Specifically: the Commander's intent, timestamps, recurrence rules, and dimension classification are all stored on-device. The AI Companion's voice response requires connectivity — but if the connection is absent, the alert still fires in text or basic audio form. AI enhances; it does not gatekeep.

For the full breakdown of what stays on-device vs what requires connectivity: Privacy Model →

Smart Escalation (Pro)

If a critical alert goes unacknowledged, SuTalk Pro notifies a pre-approved trusted contact. The user defines which events are critical and which contacts are trusted — escalation never fires without prior consent from all parties. Designed for medical, safety, and high-stakes professional use cases where missing an alert has real consequences.

The cognitive science behind why this design works →

How SuTalk compares to other AI assistants and reminder tools →