ToolsAugust 26, 2026via Vercel Blog
Gemini 3.5 Transcribe now available on AI Gateway
Why it matters
Google expands its transcription capability to the broader API ecosystem with streaming support and cost tracking, making speech-to-text accessible to practitioners building with AI Gateway's unified API layer.
Key signals
- Gemini 3.5 Transcribe now available on Google's AI Gateway
- Supports 85+ languages with auto-detection and mid-conversation language switching
- Two variants: transcribe (batch) and transcribe-live (streaming via WebSocket)
- Custom vocabulary support for names, jargon, and spellings
- AI SDK V7 adds streaming transcription via socket
- AI Gateway provides unified API, usage tracking, failover, routing rules, custom reporting, budgets
- No platform markup—reflects provider pricing only
- BYOK (Bring Your Own Key) requests supported
- Browser demo available at Gemini 3.5 Transcribe Live
- Gemini 3.5 Transcribe available on AI Gateway (unified API platform)
- Two variants: batch transcription (single request) and live streaming (WebSocket, real-time updates)
- Supports 85+ languages with automatic language detection
- Handles mid-stream language switching
- Custom vocabulary support for names, jargon, spellings
- Streaming transcription new in AI SDK V7
- AI Gateway provides usage tracking, cost monitoring, failover, performance optimization
- No platform markup or inference fees; reflects Google provider pricing
- Browser-based demo available (Gemini 3.5 Transcribe Live)
The hook
Gemini 3.5 Transcribe hits GA on AI Gateway—85+ languages, live streaming, custom vocabulary.
from Google is now available on AI Gateway. It takes audio and returns text, in two variants:Gemini 3.5 Transcribe
The model detects the language on its own, covers 85+, and follows a speaker who switches language partway through. You can also supply custom vocabulary so it recognizes names, jargon,…