Recorder is an MIT-licensed JavaScript library for HTML5 microphone recording that captures audio from any MediaStream and transcodes it into mp3, wav, ogg, webm, amr, g711a, or g711u, aimed at front-end, hybrid-app, and mini-program developers who need recording, format conversion, or speech recognition inside a web page.

What it is

Recorder lives in the browser JavaScript and web audio ecosystem. It is distributed as repository source plus a hosted demo rather than as a packaged application, and it sits directly on the getUserMedia API and on any other MediaStream object a page can obtain. The library supplies the recording and transcoding layer that a project would otherwise have to build itself: browser capture, format encoding, real-time streaming, and playback helpers all arrive in one set of JavaScript files. Its companion, RecordApp, extends the same model into WeChat mini programs and uni-app, where getUserMedia is not available at all.

The concrete problem is that browsers do not agree on what they can record. Chrome, Firefox, and Safari expose different codecs and containers, the Tencent X5 kernel used by QQ and WeChat behaves differently again, and iOS only reached getUserMedia in Safari from version 14.3 onward. Recorder normalises that situation by encoding captured buffers into a chosen output format, with mp3 as the default, and by offering real-time transcoding and transfer so audio can be sent to a server while recording continues. It also addresses memory pressure during long sessions through onProcess, which releases processed data instead of holding the entire recording.

Key capabilities

• Records from any MediaStream, including streams returned by getUserMedia, WebRTC remote streams, streams from the captureStream method on audio and video tags, and streams created by the page itself.
• Encodes to mp3 by default, with wav, pcm, g711a, g711u, ogg, amr, and beta-stage webm as alternatives, and supports custom formats when a matching encoder is supplied.
• Ships two builds with different footprints: recorder.mp3.min.js at 150kb for mp3 and recorder.wav.min.js at 25kb for wav, both with real-time transcoding and real-time transfer.
• Converts already recorded buffers or arbitrary pcm data into another format through the rec.mock method.
• Supports real-time processing and playback through plugins, including BufferStreamPlayer for segmented audio playback, audio visualisation, speed and pitch adjustment, and speech recognition.
• Handles DTMF encoding and decoding, and ships a working H5 voice call and chat example.
• Runs partly outside the browser in environments such as nodejs, and runs in WeChat mini programs and uni-app through RecordApp.

Who uses it and how

• Teams building mobile web recording inside WeChat, QQ, or a mini-program WebView, where the X5 kernel and the absence of getUserMedia in mini programs would otherwise block capture.
• Hybrid app developers who need Android and iOS builds; the repository includes Android and iOS app source, WeChat mini program source, and uni-app source under app-support-sample.
• Applications that need live transcription, using the real-time ASR tutorials that stream encoded audio toward speech recognition services as recording proceeds.
• Desktop and Electron applications that capture audio and video playback streams, using the captureStream tutorials for visualisation and processing.
• Long-running voice capture, where continuous sessions rely on onProcess to keep memory under control.

Getting started

The README describes no package manager or container image. The stated path is to download the repository source and open the html files locally, or to use the hosted copy at https://xiangyuecn.github.io/Recorder/, then include recorder.mp3.min.js or recorder.wav.min.js in the page.

How it compares

No similar tools are named in the facts, and no list of paid products it replaces is provided, so this project stands alone in this registry for the capabilities recorded here.

When to use it — and when not to

Recorder supports mono only; stereo is not adapted, which rules it out for music or multi-channel capture. Browser support has real gaps, notably older domestic Android browsers and iOS 11.0 to 14.2 outside Safari, and the repository carries 136 open issues. There is also no server component: upload handling, file storage, and playback hosting remain the integrator's responsibility, and the documentation and issue tracker are primarily in Chinese.