speechmatics-voice

What it is and what it does

Speechmatics Voice is a higher-level Python wrapper around the Speechmatics Real-Time API that simplifies building voice-enabled applications. Instead of handling raw word-level transcription events, it provides intelligent speech segmentation (grouping words into meaningful segments per speaker), automatic turn detection to know when speakers finish, and speaker management for multi-speaker scenarios. It comes with preset configurations optimized for conversational AI, voice agents, note-taking, and live captioning, so you can start with sensible defaults rather than tuning parameters from scratch.

The SDK is designed for async Python applications that stream audio and need clean, structured speech segments with speaker IDs rather than low-level transcription events. It depends on speechmatics-rt for the underlying real-time connection, pydantic for configuration validation, and numpy for audio handling. Installation is straightforward; optional ML features require additional dependencies but are gracefully disabled if not present.

Use it for:

• Build a conversational AI chatbot that detects when the user finishes speaking and responds appropriately
• Create a real-time meeting transcription tool that labels segments by speaker and formats them for note-taking
• Implement live captioning for video streams or broadcasts with automatic speaker identification
• Develop a voice agent that handles multi-turn dialogue with adaptive or ML-based turn detection
• Stream microphone audio from a web or mobile app to transcribe speech with minimal latency

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Python SDK for building real-time voice applications with automatic speech segmentation, turn detection, and speaker management on top of the Speechmatics Real-Time API.

Yes, if you are building a Python voice application that needs real-time transcription with speaker awareness and turn detection. The low install friction, active maintenance, MIT license, and preset configurations make it a practical choice. Verify that optional ML features meet your production requirements and that latency characteristics align with your use case before committing to a large deployment.

Install

speechmatics-voice on PyPI

pip install speechmatics-voice

uv add speechmatics-voice

poetry add speechmatics-voice

Installing speechmatics-voice

Before you install

Low friction installation with a pure-Python wheel. Active maintenance with recent commits; repo is not archived. Requires speechmatics-rt, pydantic, and numpy as runtime dependencies.

License in practice

MIT license permits commercial and private use with minimal restrictions; suitable for most projects.

Quickstart

pip install speechmatics-voice

import asyncio
from speechmatics.voice import VoiceAgentClient, AgentServerMessageType

client = VoiceAgentClient(api_key="your_key", preset="scribe")

@client.on(AgentServerMessageType.ADD_SEGMENT)
def on_segment(message):
    for segment in message["segments"]:
        print(f"{segment['speaker_id']}: {segment['text']}")

await client.connect()
await client.send_audio(audio_chunk)
await client.disconnect()

Requires Python 3.9 or later; optional ML-based turn detection and VAD features require additional dependencies that are gracefully disabled if not installed.

Verify before relying

• Whether ML-based turn detection (SMART_TURN) and VAD features work reliably without manual model download in production environments
• Performance characteristics and latency guarantees under high-concurrency or high-volume transcription workloads
• Compatibility with specific voice frameworks mentioned in keywords but not detailed in the excerpt

Package facts

Evidence: speechmatics_voice-0.2.8-py3-none-any.whl

Keywords: speechmatics, conversational-ai, voice, agents, websocket, real-time, pipecat, livekit

Tags

More Python Modules packages