Speechmatics iconSpeechmatics

Speechmatics offers speech-to-text and voice AI APIs for 55+ languages, with real-time and batch models, TTS, and cloud, on-prem or on-device deployment

55+ Languages
From $0.24/hr Batch Transcription List Price
$100 Free Credits, No Card

Overview

Speechmatics is a Cambridge, UK speech recognition company that sells speech-to-text APIs for voice AI, built to handle accents, overlapping speakers and multilingual conversations. One model covers 55+ languages, including code-switching when a speaker changes language mid-sentence. It offers pre-recorded (batch) and real-time transcription with several model tiers, speaker diarization, custom vocabulary, a medical model, add-ons such as translation and summaries, and text-to-speech for voice agents. Unlike most cloud-only rivals, it can run in the cloud, on-premises or on-device. Its Ursa model family launched in 2023, and Ursa 2 cut word error rate by 18% across 50 languages.

The Verdict

Who Should Use Speechmatics?

Best For

  • Teams transcribing accented or multilingual speech at scale
  • Regulated buyers that need on-prem or on-device deployment
  • Healthcare apps that need a medical-terms model

Not Ideal For

  • Voice agents where every millisecond of latency counts (one review measured it slower than Deepgram)
  • Non-developers who want a finished transcription app
  • Non-English text-to-speech

What's Great

  • High accuracy praised in G2 reviews, with low error rates in third-party tests
  • Wide language coverage with code-switching in one model
  • Cloud, on-prem and on-device deployment
  • API reviewers call easy to use and well documented
  • GDPR, HIPAA, SOC 2 Type II and ISO 27001 compliance

Watch Out For

  • Median time to final segment of 495ms versus 247ms for Deepgram in one review's test
  • TTS is English-only for now
  • Free tier allows only 2 concurrent real-time sessions
  • Custom models are Enterprise-only
  • Some G2 reviewers find pricing challenging and spend tracking weak

Pricing

View all features & details

Key Features

  • Batch and real-time speech-to-text, with real-time results in about 350ms (vendor claim)
  • Speaker diarization and custom vocabulary
  • Code-switching across languages
  • Medical model with fewer errors on key terms (vendor claim)
  • Add-ons: translation, chapters, topics, summaries, sentiment
  • Text-to-speech for voice agents
  • Discounts for opting into model training and for annual commitments

Use Cases

  • Contact center and meeting transcription
  • Speech input for voice agents
  • Media captioning and subtitling
  • Clinical documentation

Latest disclosed funding: $62M Series B led by Susquehanna Growth Equity (June 2022). Sources: Series B · Ursa · Ursa 2 · Docs

How It Compares

Feature Speechmatics Deepgram AssemblyAI
Entry batch price $0.24/hr $0.0048/min $0.15/hr
On-prem or on-device Yes See vendor See vendor
Languages 55+ See vendor See vendor

Sources: Speechmatics pricing · Deepgram page · AssemblyAI page

Is this your tool? Add a badge