@audio/encode-webm
Encode PCM audio samples to WebM (Opus) format.
WASM (libopus via libopus WASM via @audio/encode-opus) with built-in EBML/Matroska muxer — works in both node and browser.
import webm from '@audio/encode-webm';
const encoder = await webm({ sampleRate: 48000, channels: 1, bitrate: 96 });
const chunk = encoder.encode(channelData); // → Uint8Array (WebM header + cluster)
const tail = encoder.flush(); // → Uint8Array (final cluster)
// concatenate chunk + tail for complete WebM file
Options
| Option | Default | Description |
|---|---|---|
sampleRate | — | Input sample rate (required). Resampled to 48kHz internally. |
channels | 1 | 1 (mono) or 2 (stereo) |
bitrate | 64 | Target bitrate in kbps |
application | 'audio' | 'audio', 'voip', or 'lowdelay' |
Opus always encodes at 48kHz. Other rates are resampled to it as one stream by @audio/encode-opus's resampler: any chunking encodes the audio one chunk does, sample for sample.
Output uses unknown-size Segment and Clusters (streaming/live WebM), matching MediaRecorder output.
Streaming
const encoder = await webm({ sampleRate: 44100, channels: 1, bitrate: 128 });
const a = encoder.encode(chunk1); // → Uint8Array (EBML header + Segment + Tracks + first Cluster)
const b = encoder.encode(chunk2); // → Uint8Array (additional Cluster)
const c = encoder.flush(); // → Uint8Array (final Cluster with remaining samples)
// complete WebM = concat(a, b, c)
encoder.free();
License
The final block is written in a BlockGroup with DiscardPadding, so decoders that honour it (@audio/decode-webm, ffmpeg, browsers) return exactly as many samples as were encoded.