AI BriefingAssemblyAI Releases Voice Agent API
·2026.04.30 02:35
Key point
AssemblyAI has launched an API for voice agents.
1 / 2
Details
AssemblyAI released the Voice Agent API. It's an API that takes voice input and responds with voice, aimed at building production-grade voice agents.
- It offers ~1 second latency and supports real-time audio streaming input/output.
- It emphasizes recognition accuracy for practically important items such as numbers, emails, and names.
- It reflects prompt, voice, and tool updates during a call, and reduces the problem of tool calls cutting off into silence.
- Pricing is a flat $4.50 per hour, with no token-based billing and no concurrency limits.
The product adds a voice agent layer on top of AssemblyAI's existing speech-to-text, speaker detection, summarization, PII redaction, and LLM gateway features.