Skip to main content
Integrations and SDKs

Vapi integration

Learn how to integrate Speechmatics STT with Vapi.

Vapi offers native integration with Speechmatics for real-time, highly accurate speech-to-text in AI voice agents. This integration is designed for rapid deployment and handles real-world conversation complexities, such as background noise, diverse accents, and multiple speakers, with no code required.

Vapi is perfect for:

  • Task automation: Offload routine voice-based tasks
  • Live call analysis: Gain real-time insights from conversations
  • Hands-free control: Build voice-driven application interfaces
  • Global voice support: Deploy agents that speak any language
  • Interactive AI: Create sophisticated conversational bots

Features​

  • Native availability and flexible deployment: Speechmatics is embedded within the Vapi platform, simplifying the setup process.
  • High accuracy & low latency: The Speechmatics integration provides an accurate and low-latency input layer, crucial for natural conversations with AI agents.
  • Speaker diarization: Speechmatics is the only transcriber on Vapi that provides speaker diarization, which identifies and labels who said what in multi-speaker scenarios.

Quickstart​

Requirements​

UI Integration​

The Vapi UI provides a simple way to integrate Speechmatics STT into your voice AI agent.

  1. Navigate to assistants tab in the Vapi dashboard.
  2. Choose an existing assistant or create a new one.
  3. Access the transcriber tab or scroll down to the transcriber module settings.
  4. Select Speechmatics in the provider dropdown menu
  5. Optionally, you can add your Speechmatics API key, which can be generated in the portal, within the Vapi dashboard under Provider Keys

Usage​

import os
from vapi import Vapi
from vapi.types import SpeechmaticsTranscriber

client = Vapi(token=os.environ["VAPI_API_KEY"])

assistant = client.assistants.create(
    name="Speechmatics Quickstart",
    transcriber=SpeechmaticsTranscriber(
        provider="speechmatics",
        language="en",
    ),
)

print(f"Assistant ID: {assistant.id}")

Best practices​

  • Operating point: Select the enhanced model for the best accuracy in real-time voice agents.
  • Region: Choose the region closest to optimize latency.
  • Custom vocabulary: add key terms (e.g., product names, medical terminology) and leverage the "sounds like" option for tricky pronunciations.

For detailed examples, please see the Speechmatics Academy.