Speechmatics

Speechmatics is a speech-to-text platform that converts spoken audio, both pre-recorded files and live streams, into written text for use in captions, search, analytics, and voice AI applications. It supports more than 55 languages with recognition tuned for a range of accents and dialects, includes speaker diarization to separate individual speakers in a conversation, and lets customers add custom vocabulary for domain-specific terms. Originally founded in 2006 in Cambridge, UK (initially as Cantab Research) by speech recognition researcher Tony Robinson, the company now targets developers and enterprises across healthcare, media and broadcast, legal, contact centers, and meeting platforms. It offers RESTful and WebSocket APIs with SDKs for Python, JavaScript, and C#, and can be deployed in the cloud, on-premise, or on-device depending on a customer's data requirements. It holds ISO 27001, GDPR, HIPAA, and SOC 2 Type II compliance, and integrates with voice AI infrastructure tools like Vapi, Pipecat, and LiveKit, as well as Zapier and Adobe Premiere. An independent Pipecat benchmark in August 2026 reported a 1.07% pooled word error rate, the lowest among 12 services tested. An honest limitation: pricing beyond a free $100 credit tier is usage-based or fully custom, so understanding real-world cost requires talking to sales or estimating usage in advance.