Speechmatics
Speech-to-text API built for production voice agents, now with Agent STT.
About
Speechmatics is a Cambridge, UK voice AI company offering speech-to-text APIs with a long-standing focus on accuracy across accents and noisy audio. It has raised about $90M in total funding and serves transcription and real-time captioning workloads.
In September 2026 it launched Agent STT, a speech-to-text API built on its Linden model specifically for voice agents. It targets the small, high-consequence errors that derail calls (a wrong digit, a missed negation), with sub-350ms segment finalization, 55+ languages, custom vocabularies, live speaker diarization, and turn detection. In the Pipecat open STT benchmark it placed on the speed-accuracy Pareto frontier among 23 streaming models.
Limitations to note: benchmark leadership claims come from a third-party open benchmark at launch and will shift as models update; pricing is usage-based (from $0.30/hr), so high-volume voice products need cost modeling; and custom vocabulary caps and language coverage vary by tier.
What it can do
- Agent STT on the Linden model for voice agents
- 55+ languages with diarization and speaker ID
- Custom vocabulary and turn detection
- Pipecat, LiveKit, Vapi, and Telnyx integrations
Setup
Sign up for an API key and call the REST or real-time WebSocket API; SDKs and Pipecat/LiveKit integrations available.
Prerequisites
An API key; audio input pipeline for real-time or batch use.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 83; Trending score 78/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.