WebRTC enables real-time communication in browsers — including the audio capture that powers StreamTranslate's live captioning. Here's what it is and how it fits into the modern streaming stack.
Try StreamTranslate FreeWebRTC (Web Real-Time Communication) is an open standard developed by Google and the W3C that enables real-time audio, video, and data communication directly in web browsers and native applications without the need for plugins or third-party software. WebRTC powers video calls in Google Meet, Discord, Zoom (web client), and countless other real-time communication tools.
From a technical perspective, WebRTC handles peer-to-peer connection establishment (ICE/STUN/TURN), audio/video encoding and decoding (Opus, VP8/VP9, H.264), network adaptive bitrate, and packet loss concealment. The combination of these capabilities allows WebRTC sessions to maintain quality even on imperfect network connections.
For live streaming specifically, WebRTC enables ultra-low-latency delivery (under 500ms) compared to HLS's 10-30 seconds. Platforms experimenting with interactive streaming features — where viewer interactions need to feel instantaneous — use WebRTC for this reason. WHIP (WebRTC-HTTP Ingest Protocol) is an emerging standard that allows streaming software to push streams to servers using WebRTC rather than RTMP.
StreamTranslate's OBS browser source runs in an embedded Chromium environment that has access to the full Web Audio API — the browser standard for audio capture and processing. When you add StreamTranslate's browser source URL to OBS, the page requests access to your microphone (or desktop audio, depending on configuration), captures the audio stream in real time, and sends it via WebSocket to Deepgram's Nova-2 ASR servers.
This Web Audio-based capture approach has significant advantages for caption generation. It captures audio directly at the browser level without needing to intercept your RTMP stream or install audio routing software. The WebSocket connection to Deepgram maintains sub-50ms round-trip latency for audio transmission, contributing to StreamTranslate's total caption latency of under 400ms.
The browser source approach also means StreamTranslate is completely agnostic to your streaming platform and RTMP configuration. Whether you stream to Twitch, YouTube, Kick, or all three simultaneously via multistreaming, the caption system works identically because it captures audio at the source rather than from the stream output.
StreamTranslate's browser source captures your microphone directly via Web Audio APIs, no RTMP routing or audio loopback software required.
Audio is transmitted to Deepgram's ASR servers via persistent WebSocket, maintaining sub-50ms transmission latency for near-instant transcription.
Because audio is captured at the browser level rather than from RTMP output, StreamTranslate works identically regardless of where you stream to.
WebRTC is an open standard enabling real-time audio, video, and data communication in browsers without plugins. It powers video calls, live streaming, and real-time audio capture.
StreamTranslate's OBS browser source uses Web Audio APIs to capture microphone audio and stream it to Deepgram's ASR servers for live transcription via WebSocket.
WebRTC achieves under 500ms latency vs 10-30 seconds for standard HLS, making it ideal for interactive real-time streaming scenarios.
Twitch primarily uses HLS for standard streams. WebRTC is used by platforms like Discord and Zoom for ultra-low-latency interactive features.
For audio capture specifically, Web Audio APIs in the browser source provide low-latency microphone access without RTMP routing, which is how StreamTranslate captures audio for captions.