Documentation

Chat

  • Getting Started
  • Installation
  • Core Concepts
  • Escalation System
  • Widget Keys
  • Secret Keys (sk_)
  • REST API
  • Channels & Integrations
  • Messaging Platform API

Voice

  • Voice Agent
  • Setup Guide
  • Phone Numbers
  • Outbound Campaigns

Platform

  • Billing & Pricing
  • Glossary
  • FAQ

STT (Speech-to-Text)

AI & ML

Technology that converts spoken audio into written text in real-time during voice calls.

Speech-to-Text (STT) converts spoken audio into written text in real-time during voice calls. It's the first step in the voice pipeline — turning what the caller says into text that the AI can process.


Voice pipeline position
Caller Audio

→

STT

→

RAG + LLM

→

TTS

→

Audio Response

How it works

Audio frames are streamed to the speech recognition service in real-time. Two types of results are returned:

Interim results

Partial transcription as the caller speaks. Used for barge-in detection — lets the agent know someone is talking.

Final results

Complete transcription after the utterance ends. This is what gets passed to the LLM for response generation.


Configuration
  • Model — General purpose or telephony-optimized (for phone call audio quality)

  • Language — Fixed language or auto-detect

  • Endpointing — Silence duration before finalizing (200ms default)

  • Keywords — Boost recognition of domain-specific terms (e.g. your business name)

Ready to transform your customer experience?

Join businesses already using Sampan to serve customers 24/7 and boost revenue with AI.

sampan.club

AI-powered customer service for businesses.

SAMPAN TECH
202603046389 (003825593-A)

Product

© 2026 SAMPAN TECH, trading as Sampan Club (sampan.club). All rights reserved.