Profiles & voices
An agent profile has two independent settings: a response tier (how the agent thinks — speed vs. reasoning depth) and a voice (what it sounds like). For the concepts behind each, see Profiles & voices: concepts.
Choosing a response tier
Section titled “Choosing a response tier”Open a flow’s Setup stage (Advanced projects) — or the profile step for a Simple project — and pick a tier for the attached agent profile:
| Tier | What it does |
|---|---|
| Fast | Lowest latency, tuned for quick answers. The platform picks the fastest available model. |
| Standard | The balanced default between speed and reasoning depth. |
| Smart | Highest reasoning depth. Slower to first response than Fast or Standard. |
| Realtime (preview) | Selectable today, shown with a Preview badge, and billed like Smart. Until it is enabled for your environment, a profile set to this tier answers with Smart’s behavior — nothing breaks. |
Every tier uses speech recognition with automatic fallback, and honors your profile’s configured input languages.
An older profile that has no tier set (saved before tiers existed) keeps working exactly as before — nothing is silently changed.
Keyword/keyterm hints (stt_keyterms — proper nouns, product names, etc. that STT should bias
toward) work the same as before and are unaffected by which tier you pick.
Choosing a voice
Section titled “Choosing a voice”The output-voice picker lists every voice registered for your environment, each in one of three price tiers:
| Voice tier | Bills at | Voices | Plan required |
|---|---|---|---|
| Standard | the base per-minute rate | Nairon voices | any plan |
| Premium | ×1.3 | premium voices | Growth or above |
| HighRes | ×1.5 | high-fidelity voices | Professional or above |
A voice above your plan’s tier is shown locked in the picker rather than silently swapped for another one. Pick any voice for any profile — voice and response tier are independent choices; the two price multipliers combine (see Profiles & voices: concepts).
If a call turn asks the agent to read out an email address or similar text aloud, the agent can
speak it in a natural, space-delimited way (e.g. “support at nairon punkt a i” for
support@nairon.ai) — coming soon.