AI Signal 574 2 feeds carried it
Gemini 3.5 Transcribe now available on AI Gateway
↗
Vercel's AI Gateway now routes Google's Gemini 3.5 Transcribe in two forms, a one-shot model for complete audio files and a WebSocket-backed model that streams partial transcripts while audio is still arriving, with automatic language detection across 85+ languages.
Engineers get a single gateway endpoint for both batch and live transcription, so cost tracking, failover and key management collapse into one place rather than splitting across providers. The live variant accepts a raw ReadableStream of PCM chunks, so a microphone can be piped straight in, but the streaming API is marked experimental and pinned to AI SDK V7. The adoption cost is an SDK upgrade plus conformance to the 16 kHz 16-bit PCM format on whatever audio source you wire up; what you give up is API stability until the experimental prefix is dropped.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Two model IDs are exposed: google/gemini-3.5-transcribe for one-shot transcription of a complete file and google/gemini-3.5-transcribe-live for WebSocket streaming that returns partial transcripts while audio is still arriving.
The model auto-detects among 85+ languages and follows a speaker who switches language mid-conversation, with custom vocabulary for names and jargon, and a providerOptions.google.mode toggle between the default VERBATIM and SMART transcription modes.
Streaming uses the new experimental_streamTranscribe function in AI SDK V7 and takes a ReadableStream of audio chunks, while AI Gateway charges no platform markup on inference and supports BYOK billing.
THE CLUSTER