Pure-WASM audio decoders. No native addons, no FFmpeg, no external JS dependencies. Every decoder is a hand-compiled C library → WASM using Emscripten.

📥 Install

npm install @sonata-sdk/decoder

🚀 Usage

import { detectFormat, decodeMP3, decodeFLAC } from '@sonata-sdk/decoder'
import { readFileSync } from 'fs'

const data = readFileSync('song.mp3')
const fmt = detectFormat(data) // 'mp3' | 'flac' | 'aac' | null

const { channelData, sampleRate, samplesDecoded } = fmt === 'mp3'
  ? await decodeMP3(data)
  : await decodeFLAC(data)

// channelData[0] = Float32Array left
// channelData[1] = Float32Array right

Streaming

import { createMP3Decoder } from '@sonata-sdk/decoder'

const decoder = await createMP3Decoder()
for (const chunk of chunks) {
  const { channelData, sampleRate } = await decoder.decode(chunk)
  // process float PCM...
}
decoder.free()

Subpath imports

import { createAACDecoder } from '@sonata-sdk/decoder/aac'
import { createMP3Decoder } from '@sonata-sdk/decoder/mp3'
import { createFLACDecoder } from '@sonata-sdk/decoder/flac'

📦 What's inside

| Format | Decoder | WASM size | |--------|---------|-----------| | MP3 | mpg123 1.32.5 compiled to WASM | ~235 KB | | FLAC | libFLAC 1.4.3 compiled to WASM | ~59 KB | | AAC | FAAD2 compiled to WASM | ~276 KB |

All WASM binaries are compiled from source with Emscripten 3.1.56 and bundled in the package. Zero runtime dependencies — no @audio/decode-*, no mpg123-decoder, no @wasm-audio-decoders/.

The WASM uses a handle-based API so multiple decoders can run concurrently.

📦 Related

• Sonata — Lavalink-compatible audio server
• @sonata-sdk/voice — Discord voice connection
• @sonata-sdk/ws — Resumable WebSocket client
• sonata-sdk-packages — Monorepo