Speechmatics Launches Agent STT to Fix Critical Voice Agent Errors
Event summary
- Speechmatics launched Agent STT, a speech-to-text API designed for voice agents, powered by its new Linden model.
- Agent STT focuses on semantic accuracy to catch high-consequence errors in voice agent interactions.
- Linden model recorded a 1.05% pooled semantic error rate with a 369ms median finalization time in Pipecat benchmark.
- Agent STT is available now, priced from $0.30 per hour with volume discounts down to $0.16 per hour.
The big picture
Speechmatics' Agent STT addresses a critical gap in voice AI, where small transcription errors can derail entire interactions. This launch comes as enterprises increasingly rely on voice agents for customer service and operational efficiency, making semantic accuracy a key differentiator. The integration with platforms like Pipecat and LiveKit simplifies adoption, potentially accelerating the deployment of more reliable voice agents across industries.
What we're watching
- Adoption Pace
- How quickly enterprises will integrate Agent STT into their voice agent stacks.
- Competitive Edge
- Whether Speechmatics can sustain its lead in semantic accuracy against competitors.
- Market Expansion
- The pace at which Speechmatics expands its 55+ language support to capture global markets.
Related topics