Speechmatics

Speechmatics is a business-focused voice AI platform founded in 2006 and headquartered in Cambridge, England. It provides speech-to-text, text-to-speech and voice agent technologies through cloud APIs and deployable software.

The platform supports real-time and batch transcription across more than 55 languages. Its documented capabilities include speaker and channel diarization, word-level timestamps, custom dictionaries, confidence scores, translation, summaries, sentiment analysis and topic detection.

Speechmatics supports cloud, on-premises and on-device deployments. On-premises options include Docker containers, Kubernetes and a virtual appliance. On-device deployment is documented for Mac and Windows environments.

Speechmatics provides real-time transcription through streaming connections and batch transcription for uploaded audio or video files. Real-time transcripts can begin arriving within milliseconds, while final output improves as additional context becomes available.

• Speech-to-text for live streams and prerecorded media
• Support for more than 55 languages and dialects
• Speaker and channel diarization
• Word-level timestamps and confidence scores
• Custom dictionaries for technical terms, acronyms and proper nouns
• Automatic language identification and multilingual processing
• Translation to and from English for more than 30 languages
• Numeral formatting, punctuation, casing, profanity and disfluency handling
• Summaries, sentiment analysis, topics and chapters
• REST and WebSocket APIs
• Python and Node.js SDKs
• Integrations with LiveKit, Pipecat, Vapi and NVIDIA Holoscan

Speechmatics offers a free entry tier with $100 in usage credit and no credit card requirement. The free tier includes speech-to-text access, 55-plus languages, two concurrent real-time sessions and multi-region cloud options.

The Pro tier starts at $0.129 per hour. It includes 50 concurrent real-time sessions, up to 10 file jobs per second, online email support and the same initial $100 credit. Speechmatics also advertises a 20% discount option.

Enterprise pricing is quote-based. It adds volume discounts, custom models, unlimited scale, custom language development, SaaS or on-premises deployment, higher concurrency and prioritized support. Pricing can vary by model, feature and deployment.

Pros

• Supports both real-time and batch transcription workflows.
• Offers broad multilingual coverage with speaker diarization.
• Provides cloud, on-premises and on-device deployment choices.
• Includes documented REST, WebSocket and SDK-based integration paths.
• Supports technical customization through custom dictionaries and metadata.

Cons

• Enterprise deployment and custom models require a sales process.
• Feature and deployment costs can vary beyond the base hourly rate.
• The product is primarily designed for developers, platforms and enterprises.
• On-device deployment is documented for Mac and Windows, limiting stated desktop coverage.

Comparable speech-to-text and voice AI services include Deepgram, AssemblyAI, Google Cloud Speech-to-Text, Amazon Transcribe, Microsoft Azure AI Speech, OpenAI speech-to-text and ElevenLabs Scribe. Speechmatics is particularly relevant when multilingual support, diarization and controlled deployment are priorities.

Does Speechmatics offer a free trial?

Yes. The free tier provides $100 in usage credit without requiring a credit card. The pricing page describes this as free access for developers and early exploration.

Does Speechmatics have an API?

Yes. Speechmatics documents REST and WebSocket APIs for batch and real-time transcription. It also provides Python and Node.js SDKs.

What integrations does Speechmatics support?

Official documentation lists native integrations with LiveKit, Pipecat, Vapi and NVIDIA Holoscan. The platform also provides SDKs and API references for custom integrations.

Which platforms support Speechmatics?

Speechmatics is available through cloud APIs and can be deployed on-premises using Docker, Kubernetes or a virtual appliance. Its on-device documentation covers Mac and Windows.

How many languages does Speechmatics support?

Speechmatics advertises support for more than 55 languages. Its Melia model is designed for multilingual audio and code-switching across supported languages.

Can Speechmatics identify speakers?

Yes. Speaker and channel diarization are available for batch and real-time transcription. The company states that real-time streams can identify up to 50 speakers by default, with higher limits available.