Sign InOpen Brain
VercelEngineering PostOfficial Source

AI Gateway now supports streaming transcription

AI Gateway can now stream audio into transcription models and emit partial text, letting text-based agents accept lower-latency voice input without changing the agent itself.

Vercel · Jul 22, 2026
Open Source Open MarkdownOpen JSON
Source Summary

AI Gateway now supports **streaming transcription** through the AI SDK’s **streamTranscribe** function. It accepts live audio and emits transcript deltas plus partial and final transcripts across compatible providers.

Practical Implication

Builders can add voice input by forwarding the live transcript to an existing text agent; swap the model string to change providers, and add speech generation only when spoken output is required.

Agent-Ready Context
AI Gateway now supports **streaming transcription** through the AI SDK’s **streamTranscribe** function. It accepts live audio and emits transcript deltas plus partial and final transcripts across compatible providers.

Builders can add voice input by forwarding the live transcript to an existing text agent; swap the model string to change providers, and add speech generation only when spoken output is required.

The feature is **beta**, and the material gives no latency, accuracy, interruption-handling, or cost measurements. Full two-way conversation may be better served by a realtime voice system.
Context Map
infraaudio#gateways#tool-use#agent-sdks
Uncertainty
The feature is **beta**, and the material gives no latency, accuracy, interruption-handling, or cost measurements. Full two-way conversation may be better served by a realtime voice system.