Summary
On July 22, 2026, Vercel added streaming transcription to AI Gateway, letting developers stream live audio and receive transcript updates as they are produced through the AI SDK's streamTranscribe. It extends the Gateway's earlier realtime voice, speech-to-text, and text-to-speech support toward lower-latency voice agent workflows.
What changed
Vercel AI Gateway now supports streaming transcription via the AI SDK's streamTranscribe, returning incremental transcript updates for live audio instead of only post-hoc results.
Why it matters
Incremental transcription lets voice agents and live-captioning apps react before a speaker finishes, cutting perceived latency and enabling barge-in and mid-utterance tool calls, and keeps Vercel's AI Gateway competitive as a full realtime-voice provider rather than a text-only model router.
Evidence excerpt
Stream live audio and get transcript updates back as they're produced, using the AI SDK's streamTranscribe.