Realtime API supports WebSocket, WebRTC, and AOQ transport protocols
Realtime API provides three transport protocols -- WebSocket, WebRTC, and AOQ (AI over QUIC) -- each optimized for different latency, platform, and integration requirements.
Realtime API supports three transport protocols: WebSocket, WebRTC, and AOQ (AI over QUIC). Choose the protocol that best fits your use case.
Choose a protocol based on your requirements:
The following table shows which models and application types each protocol supports.
Overview
Realtime API supports three transport protocols: WebSocket, WebRTC, and AOQ (AI over QUIC). Choose the protocol that best fits your use case.
| Dimension | WebSocket | WebRTC | AOQ |
|---|---|---|---|
| Recommended use cases | Server-side integration, rapid prototyping | Browser-based interaction, traditional audio/video calls | AI multimodal real-time interaction, poor network conditions, multimodal data transport |
| Browser compatibility | Native support | Native support | Not supported |
| Integration complexity | Very low | Medium | Low |
| Poor network resilience | Poor | Good | Exceptional |
| Data types | Text / audio / image | Audio/video + text | Audio/video + text |
| Connection speed | Slow | Slow | Fast |
| Echo cancellation / noise reduction | None -- handled client-side | Built-in | Built-in |
| AI suitability | Basic -- suited for text-only or latency-tolerant use cases | Designed for traditional audio/video; requires extra adaptation for AI | Purpose-built for AI multimodal workloads |
| Client platform support | All platforms (any environment with WebSocket support) | Browser, mobile | Android / iOS / HarmonyOS |
- WebSocket: Best for server-side integration, rapid prototyping, and scenarios where integration simplicity matters most. Use the DashScope SDK to get a real-time voice conversation running quickly.
- WebRTC: Best for browser-native scenarios and applications already built on WebRTC infrastructure, such as traditional audio/video calling. Includes built-in echo cancellation and noise reduction.
- AOQ: Best for AI real-time interaction with strict latency, poor network resilience, and multimodal data transport requirements -- especially native mobile apps. Includes built-in echo cancellation and noise reduction.
Model and application support
The following table shows which models and application types each protocol supports.
| Application type | Model | AOQ | WebRTC | WebSocket |
|---|---|---|---|---|
| Real-time omni | qwen3.5-omni-plus-realtime | Supported | Supported | Supported |
| Real-time omni | qwen3.5-omni-flash-realtime | Supported | Supported | Supported |
| Real-time omni | qwen3.5-livetranslate-flash-realtime | Supported | Supported | Supported |
| Real-time speech recognition | FunASR series models | Unsupported | Unsupported | Supported |
| Real-time speech synthesis | CosyVoice series models | Unsupported | Unsupported | Supported |
For model names, context windows, pricing, and snapshot versions, see the QwenCloud console. For concurrency limits, see Rate limiting.