Communication
WhatsApp voice messages
How OrderInChat turns customer voice notes into the same AI brain as text — STT, reply, TTS (≤90s).
Overview
Customers can send voice notes on WhatsApp. OrderInChat does not use a separate “voice-only brain.” The path is:
- STT — transcribe the note (speech-to-text)
- Same support agent — the transcript is handled like a normal WhatsApp text message (tools, catalog, booking, checkout rules)
- TTS — speak the reply back as audio (text-to-speech), capped at about 90 seconds
Configure models under Assistant → AI Models (STT, TTS, and the customer chat model). Live / Realtime voices are optional and separate from this pipeline.
What merchants should expect
- Replies use your AI Instructions, Knowledge Base, and live catalog/booking data — same quality bar as typed chat
- Long notes may take longer to transcribe; OrderInChat allows enough time for both transcription and the reply, so customers get a real answer instead of an apology message
- Catalog “live reading” while speaking is off; the AI answers from your catalog and Knowledge Base after the note is transcribed
- Outside Meta’s 24-hour window, free-form replies may still require an approved template (same as text)
Setup checklist
- WhatsApp Business connected (Channels & integrations → WhatsApp Business)
- AI Models: STT + TTS + support chat model selected
- AI Instructions and Knowledge Base ready for your catalog/support facts
- Send a test voice note from a customer number and confirm you get a spoken reply with correct product/price facts
