Speechmatics
Speech-to-text API pitched at accents, noise and mixed-language audio
Overview
Speechmatics is a Cambridge, UK speech-to-text API for pre-recorded and live audio in 55+ languages, with diarisation, translation and summaries as add-ons, plus text-to-speech. It prices per audio hour by model, starts new accounts with a $100 credit, and offers on-premises and on-device deployment on Enterprise contracts.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 2 concurrent real-time sessions
- All self-serve models
- 55+ languages
- Speaker diarization and custom dictionary
- 10 file jobs per second on self-serve
- 50 concurrent sessions on self-serve
- Turn detection on Linden 1
- Low-latency speech synthesis for voice AI
- Dedicated CSM and solutions engineer
- Containers, virtual appliance, on-device and hybrid deployment
- Early feature access
Use cases
What you can produce with Speechmatics
- Transcripts of recorded files and live audio in 55+ languages
- Mixed-language transcripts where speakers switch languages mid-sentence (Melia)
- Speaker-labelled transcripts with word-level timings
- Low-latency transcription for voice agents with turn detection (Linden)
- Translation, summaries, sentiment, topics and chapters as paid add-ons
- Synthesised speech from text through the TTS API
- On-premises, on-device or hybrid deployment under an Enterprise contract
Speechmatics is serious speech infrastructure for teams that will call an API: published per-hour prices by model, real-time and batch transcription, code-switching across languages and a voice-agent model built for fast turns. Its accent and accuracy claims are its own benchmarks, so test on your own audio.
The $100 starter credit is enough for that test. On-premises and on-device deployment, the reason broadcasters and public-sector buyers pick it, require an Enterprise contract.
What we found
Speechmatics is a Cambridge, UK speech recognition company, trading through Cantab Research Ltd, which was incorporated in 2006. Its founder, Tony Robinson, worked on applying neural networks to speech recognition at Cambridge University in the 1980s. The company raised a $62 million Series B led by Susquehanna Growth Equity in 2022; AlbionVC and IQ Capital are earlier backers.
The product is an API, not an app. It transcribes pre-recorded files and live audio in more than 55 languages, with speaker diarisation, custom dictionaries and word-level timings, plus paid add-ons for translation, summaries, sentiment, topics and chapters. Speechmatics' own marketing leans on accents, noisy audio and speakers switching languages mid-sentence. Its 2026 models push that line: Melia, launched in June in production preview for batch transcription, handles code-switching across all supported languages, and Linden, launched in September for voice agents, targets fast turn-taking and spoken digits and letters. The benchmarks quoted for both, against Deepgram, Microsoft and AssemblyAI, are Speechmatics' own. The older Flow voice-agent product name has gone; its page now redirects to voice agents.
Pricing is per hour of audio and published by model. Pre-recorded transcription runs from $0.12 an hour on Melia to $0.38 on Enhanced, the most accurate model; real-time Enhanced costs $0.40, and Linden is listed at a reduced $0.15. Volume discounts reach 25%. New accounts start with a one-off $100 credit and no card, not a monthly free allowance. Text-to-speech costs $0.011 per 1,000 characters after the first million.
The constraint is deployment. Self-serve accounts get cloud processing in the EU, US or Australia and 50 real-time sessions; containers, a virtual appliance, on-device and hybrid deployment are Enterprise only. Speechmatics lists SOC 2 Type II, ISO/IEC 27001:2022, GDPR and HIPAA compliance and says customer audio is not logged for model training by default. ASEAN teams should test their own languages and accents with the starter credit before committing.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Speechmatics's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Speechmatics unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Speechmatics directly →
Spotted something out of date? Suggest an update →
Alternatives to Speechmatics
More in Video & Audio
