# What are the main benefits of migrating to Voice API v2?

Voice API v2 is a ground-up redesign that gives developers more control, better AI integration, and a cleaner architecture for building modern voice applications. Here’s what’s new:

## Cleaner call model with Sessions and Bridges

V2 introduces a Session/Bridge architecture that replaces the flat call-leg model in V1. A session groups all related calls into a single context, making it significantly easier to manage multi-party scenarios, call transfers, and parallel legs without stitching together separate API requests.

## Native AI voice integration — no glue code required

V2 ships with two dedicated AI integration paths built directly into the API:

- **Voice Relay** — a WebSocket bridge that streams real-time speech-to-text to your AI agent and plays back text-to-speech responses with barge-in support and interruption detection. Connect to any LLM or AI agent in real time.
- **Raw audio streaming** — pipe raw PCM audio directly to any WebSocket endpoint if you prefer to bring your own STT/TTS stack.


Combined with built-in Answering Machine Detection (AMD) — which classifies calls as human, machine, beep, or unknown, each with its own event handler — v2 covers the full AI voice application lifecycle out of the box.

## 500+ TTS voices with SSML support

The expanded voice library gives you over 500 voices including AI voices across multitude of languages and accents. All text-to-speech commands support SSML for fine-grained control over pronunciation, pacing, and emphasis.

## Richer call control with SVAML v2

The SVAML command language has been significantly expanded. New commands include startRecording and stopRecording with direct delivery to AWS S3, Google Cloud Storage, or Azure Blob — including optional transcription. The full command set is schema-validated, making it reliable for both developers and AI-assisted code generation.

## Batch calling with intelligent call pacing

New batch endpoints let you run large outbound campaigns with configurable call pacing, TTL controls, and per-call parameter interpolation — without managing your own queue infrastructure.

## Unified authentication across Sinch products

V2 supports both OAuth 2.0 and Basic Auth, with a single credential set that works across Sinch products. Six named regional endpoints (US, EU, Brazil, Singapore, Australia, and a global redirect) give you latency control for global deployments.

## CloudEvents-compliant webhooks

Webhook events follow the CloudEvents 1.0 standard, making them directly compatible with event-driven infrastructure such as AWS EventBridge, Apache Kafka, and similar platforms — with no custom parsing required.