Speechmatics STT integration guide

Available in

Python

Overview

Speechmatics provides enterprise-grade speech-to-text APIs. Their advanced speech models deliver highly accurate transcriptions across diverse languages, dialects, and accents. You can use the LiveKit Speechmatics plugin with the Agents framework to build voice AI agents that provide reliable, realtime transcriptions.

Quick reference

This section includes a basic usage example and some reference material. For links to more detailed documentation, see Additional resources.

Installation

Install the plugin from PyPI:

pip install "livekit-agents[speechmatics]~=1.2"

Authentication

The Speechmatics plugin requires an API key.

Set SPEECHMATICS_API_KEY in your .env file.

Usage

Use Speechmatics STT in an AgentSession or as a standalone transcription service. For example, you can use this STT in the Voice AI quickstart.

from livekit.plugins import speechmatics

session = AgentSession(
   stt = speechmatics.STT(),
   # ... llm, tts, etc.
)

Speaker diarization

You can enable speaker diarization to identify individual speakers and their speech. When enabled, the transcription output can include this information through the speaker_id and text attributes.

See the following for example configurations and outputs:

<{speaker_id}>{text}</{speaker_id}>: <S1>Hello</S1>.
[Speaker {speaker_id}] {text}: [Speaker S1] Hello.

stt = speechmatics.STT(
   enable_diarization=True,
   speaker_active_format="<{speaker_id}>{text}</{speaker_id}>",
)

Inform the LLM of the format for speaker identification by including it in your agent instructions. For a an example, see the following:

Speechmatics STT speaker diarization

An example of using Speechmatics to identify speakers in a multi-speaker conversation.

Parameters

This section describes the key parameters for the Speechmatics STT plugin. See the plugin reference for a complete list of all available parameters.

operating_pointstringOptionalDefault: enhanced

Operating point to use for the transcription. This parameter balances accuracy, speed, and resource usage. To learn more, see Operating points.

languagestringOptionalDefault: en

ISO 639-1 language code. All languages are global, meaning that regardless of which language you select, the system can recognize different dialects and accents. To see the full list, see Supported Languages.

enable_partialsboolOptionalDefault: true

Enable partial transcripts. Partial transcripts allow you to receive preliminary transcriptions and update as more context is available until the higher-accuracy final transcript is returned. Partials are returned faster but without any post-processing such as formatting. When enabled, the STT service emits INTERIM_TRANSCRIPT events.

enable_diarizationboolOptionalDefault: false

Enable speaker diarization. When enabled, spoken words are attributed to unique speakers. You can use the speaker_sensitivity parameter to adjust the sensitivity of diarization. To learn more, see Diarization.

max_delaynumberOptionalDefault: 1.0

The maximum delay in seconds between the end of a spoken word and returning the final transcript results. Lower values can have an impact on accuracy.

end_of_utterance_silence_triggerfloatOptionalDefault: 0.5

The maximum delay in seconds of silence after the end of turn before the STT service returns the final transcript.

end_of_utterance_modeEndOfUtteranceModeOptionalDefault: EndOfUtteranceMode.FIXED

The delay mode to use for triggering end of turn. Valid values are:

EndOfUtteranceMode.FIXED: Delay is fixed to the value of end_of_utterance_silence_trigger.
EndOfUtteranceMode.ADAPTIVE: Delay can be adjusted by the content of what the most recent speaker has said, including rate of speech and speaking patterns (for example, pauses).
EndOfUtteranceMode.NONE: Disables end of turn detection and uses a fallback timer.

To use LiveKit's end of turn detector model, set this parameter to EndOfUtteranceMode.NONE.

speaker_active_formatstringOptional

Formatter for speaker identification in transcription output. The following attributes are available:

{speaker_id}: The ID of the speaker.
{text}: The text spoken by the speaker.

By default, if speaker diarization is enabled and this parameter is not set, the transcription output is not formatted for speaker identification.

The system instructions for the language model might need to include any necessary instructions to handle the formatting. To learn more, see Speaker diarization.

diarization_sensitivityfloatOptionalDefault: 0.5

Sensitivity of speaker detection. Valid values are between 0 and 1. Higher values increase sensitivity and can help when two or more speakers have similar voices. To learn more, see Speaker sensitivity.

The enable_diarization parameter must be set to True for this parameter to take effect.

prefer_current_speakerboolOptionalDefault: false

When speaker diarization is enabled and this is set to True, it reduces the likelihood of switching between similar sounding speakers. To learn more, see Prefer current speaker.

Additional resources

The following resources provide more information about using Speechmatics with LiveKit Agents.

Python package

The livekit-plugins-speechmatics package on PyPI.

Plugin reference

Reference for the Speechmatics STT plugin.

GitHub repo

View the source or contribute to the LiveKit Speechmatics STT plugin.

Speechmatics docs

Speechmatics STT docs.

Voice AI quickstart

Get started with LiveKit Agents and Speechmatics STT.