Build phone and voice experiences fast—answer calls, read text aloud, or transcribe speech for your workflows.

Costs are shown in credits in the UI (1 credit = $0.01). Sonic voice is ~$0.24/min, TTS is ~$0.36 per 1M chars, STT is ~0.72 credits/min.

A) Live Voice Hotline (Sonic Realtime)

In Twilio, point your voice number webhook to the Nuxa voice endpoint (from Settings → Voice). This forwards calls into your workflow.

Use the Start block (call audio) → Agent → Response. Enable the Sonic/Realtime voice option so the agent can talk back during the call.

Attach your knowledge base (pricing, shipping, policies) so answers stay factual. Add a Guardrail block to prevent collecting sensitive data.

In the Agent prompt, say “If confidence is low, offer a callback and summarize the ask.” Log the summary to Slack/Email.

Result: A 24/7 voice agent that answers FAQs and hands off when needed.

B) Text-to-Speech (TTS)

Use case: Confirm orders, read status updates, or narrate summaries.
Blocks: Start → Agent (or Function) → Response.
Setup: Enable TTS on the Response block; choose your voice provider (e.g., ElevenLabs).
Tip: Keep messages short; add customer name/order ID for a personal touch.

Use case: Voicemail-to-text, meeting notes, intake forms via voice.
Blocks: Start (audio input) → STT option → Agent/Function.
Setup: Enable STT on the Start or Agent tool; choose provider (Whisper/Deepgram/etc.).
Tip: After transcription, run a short Agent prompt to normalize fields (name, reason, callback).

FAQ hotline: Start → Agent (KB) → Response (voice) → Slack log.
Call → Ticket: Start → STT → Agent → API (Zendesk/HubSpot) → SMS confirmation.
Outbound reminders: Schedule trigger → Agent writes script → Response (TTS) places call via Twilio.

If audio is choppy, lower model verbosity and keep responses concise.
If answers drift, make the system prompt: “Answer only from the knowledge base; if missing, ask 1 clarifying question.”
Monitor credits in Settings → Subscription—voice minutes accumulate faster than chat.