Beta Vercel

Vercel AI SDK realtime (via AI Gateway)

Beta realtime voice support in the AI SDK through Vercel AI Gateway: gateway.experimental_realtime mints connection tokens server-side and a useRealtime React hook handles mic capture and playback for speech-to-speech models.

Est. per minuten/a

Overview

Best for: Next.js/React apps already using the AI SDK and AI Gateway that want a quick browser voice mode.

At a glance

Platform $/min$0

All-in priceNo

Entry plan $/mo$0

Free tierYes

Phone numbersNo

WebRTCYes

WebSocketYes

Self-hostNo

Open weightsYes

Beta/canary API. SDK is free; AI Gateway passes through model cost (markup on realtime models unverified).

Audio in

Browser mic

Audio out

Browser playback

Languages

Per model.

Latency

Model-dependent.

Regions

Vercel AI Gateway.

Compliance

Not covered.

Features

one gateway key for multiple realtime model vendors
React hook
batch STT/TTS via AI Gateway

Pricing

What	Price	Unit	Notes
AI SDK	$0	license	AI Gateway passes through model usage; see Vercel AI Gateway pricing and model segment.

How the per-minute estimate was worked out

Equals the underlying realtime model cost; see models segment.

Free tier: SDK open source; AI Gateway has its own terms.

Source: vercel.com

Setup

pnpm add ai@canary @ai-sdk/gateway@canary @ai-sdk/react@canary (or AI SDK 7 per changelog).
Create a server route that uses gateway.experimental_realtime getToken to mint a connection.
Use the useRealtime hook in a React client for mic and playback.

Endpoint

Vercel AI Gateway

Authentication

AI_GATEWAY_API_KEY or Vercel OIDC

Quick start javascript

// Vercel AI SDK realtime is Beta and its signatures are changing; follow
// https://vercel.com/docs/ai-gateway/modalities/realtime for current code.
// Install: pnpm add ai@canary @ai-sdk/gateway@canary @ai-sdk/react@canary
// Server: gateway.experimental_realtime -> getToken() for a model such as "openai/gpt-realtime-2"
// Client: const rt = useRealtime({ ...token endpoint... }) -> start/stop mic + playback

Written from the current docs. Check the vendor's SDK version before you ship.

Warnings

Canary/Beta API

Exact function signatures were not stable enough to publish a full snippet; expect breaking changes.

Version confusion

Docs say canary releases; changelog says AI SDK 7. Use whichever the docs page you follow specifies.

Speech-to-speech only models

Some gateway realtime models do not support transcription, so you cannot get user transcripts from them.

Plus 12 warnings that apply to all platforms and telephony APIs. See category warnings.

Limits

Beta / canary API surface
Some models are speech-to-speech only (no transcription)

Models and products

Name	Status	Notes
gateway.experimental_realtime + useRealtime	Beta	Changelog (29 Jun 2026) says available via AI SDK 7; docs say install canary packages. Examples use openai/gpt-realtime-2 and xai/grok-voice-think-fast-1.0.

Docs and sources

Docs

Sources used

Not fully verified

Exact API signatures; gateway markup on realtime models.

Similar platforms APIs

Agora / TEN communityTEN Frameworkn/a0 high-severity warnings Vonage (Ericsson)Vonage Voice/Video API AI connectors (Audio Connector) + AI Studion/a0 high-severity warnings OpenAIOpenAI Agents SDK (voice / realtime agents)n/a0 high-severity warnings VoiceflowVoiceflow (voice agents)n/a0 high-severity warnings

Spotted a wrong price or a dead link?