@ai-ecoverse/pi-audio 1.2.0 → 1.3.0

package/README.md

@@ -21,6 +21,8 @@ addAudioCounters(total, usage);
 
 pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatches in `patches/patches.json`.
 
+`convertAudio` and `convertAudioSource` convert between `AudioFormat`s: PCM16, float32, G.711 μ-law and A-law, mono and stereo, sample rate, and PCM16 WAV. Opus, MP3 and FLAC pass through only when the format does not change. `encodeAudioBase64` and `decodeAudioBase64` are the wire helpers for JSON protocols. `AudioChunk.data` stays binary. The same converter runs in an AudioWorklet from `@ai-ecoverse/pi-audio/worklet` (`AUDIO_CONVERT_PROCESSOR`, `registerAudioConvertWorklet`). Adapters declare `inputFormats` and `outputFormats` and do not resample on their own.
+
 ## Development
 
 ```sh

package/dist/convert.js

@@ -0,0 +1,294 @@
+const compressed = /* @__PURE__ */ new Set(["opus", "mp3", "flac"]);
+const LAW_BIAS = 132;
+const base64Alphabet = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
+function byteAt(data, index) {
+  const byte = data[index];
+  return byte === void 0 ? 0 : byte;
+}
+function symbolAt(index) {
+  return base64Alphabet.charAt(index);
+}
+function encodeAudioBase64(data) {
+  let encoded = "";
+  for (let index = 0; index < data.length; index += 3) {
+    const second = data[index + 1];
+    const third = data[index + 2];
+    const triple = byteAt(data, index) << 16 | byteAt(data, index + 1) << 8 | byteAt(data, index + 2);
+    encoded += symbolAt(triple >> 18 & 63);
+    encoded += symbolAt(triple >> 12 & 63);
+    encoded += second === void 0 ? "=" : symbolAt(triple >> 6 & 63);
+    encoded += third === void 0 ? "=" : symbolAt(triple & 63);
+  }
+  return encoded;
+}
+function codeAt(value, index) {
+  const character = value[index];
+  if (character === void 0) return -1;
+  const code = base64Alphabet.indexOf(character);
+  if (code < 0) throw new Error("Invalid base64 audio");
+  return code;
+}
+function sextet(code) {
+  return code < 0 ? 0 : code;
+}
+function decodeAudioBase64(value) {
+  const cleaned = value.replaceAll("=", "");
+  if (cleaned.length % 4 === 1) throw new Error("Invalid base64 audio");
+  const bytes = [];
+  for (let index = 0; index < cleaned.length; index += 4) {
+    const first = codeAt(cleaned, index);
+    const second = codeAt(cleaned, index + 1);
+    const third = codeAt(cleaned, index + 2);
+    const fourth = codeAt(cleaned, index + 3);
+    const triple = sextet(first) << 18 | sextet(second) << 12 | sextet(third) << 6 | sextet(fourth);
+    bytes.push(triple >> 16 & 255);
+    if (third >= 0) bytes.push(triple >> 8 & 255);
+    if (fourth >= 0) bytes.push(triple & 255);
+  }
+  return Uint8Array.from(bytes);
+}
+function convertAudio(chunk, format) {
+  checkFormat(chunk.format);
+  checkFormat(format);
+  if (sameFormat(chunk.format, format)) {
+    return { data: chunk.data.slice(), format, timestamp: chunk.timestamp };
+  }
+  if (compressed.has(chunk.format.encoding) || compressed.has(format.encoding)) {
+    throw new Error(
+      `Cannot convert ${chunk.format.encoding} to ${format.encoding} without a codec`
+    );
+  }
+  const decoded = decode(chunk.data, chunk.format);
+  const mixed = mix(decoded.channels, format.channels);
+  const resampled = mixed.map(
+    (channel) => resample(channel, decoded.sampleRate, format.sampleRate)
+  );
+  return {
+    data: encode(resampled, format),
+    format,
+    timestamp: chunk.timestamp
+  };
+}
+async function* convertAudioSource(source, format) {
+  for await (const chunk of source) yield convertAudio(chunk, format);
+}
+function sameFormat(left, right) {
+  return left.encoding === right.encoding && left.sampleRate === right.sampleRate && left.channels === right.channels;
+}
+function checkFormat(format) {
+  if (format.channels !== 1 && format.channels !== 2) {
+    throw new Error(`Unsupported channel count ${String(format.channels)}`);
+  }
+  if (!Number.isFinite(format.sampleRate) || format.sampleRate <= 0) {
+    throw new Error(`Unsupported sample rate ${String(format.sampleRate)}`);
+  }
+}
+function decode(data, format) {
+  if (format.encoding === "wav") return decodeWav(data);
+  return { channels: decodePcm(data, format), sampleRate: format.sampleRate };
+}
+function decodePcm(data, format) {
+  const width = sampleWidth(format.encoding);
+  const frames = Math.floor(data.byteLength / (width * format.channels));
+  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);
+  const channels = Array.from({ length: format.channels }, () => new Float32Array(frames));
+  for (let frame = 0; frame < frames; frame++) {
+    for (let channel = 0; channel < format.channels; channel++) {
+      const offset = (frame * format.channels + channel) * width;
+      channels[channel][frame] = readSample(view, offset, format.encoding);
+    }
+  }
+  return channels;
+}
+function sampleWidth(encoding) {
+  if (encoding === "pcm16") return 2;
+  if (encoding === "pcmf32") return 4;
+  if (encoding === "mulaw" || encoding === "alaw") return 1;
+  throw new Error(`Unsupported audio encoding ${encoding}`);
+}
+function readSample(view, offset, encoding) {
+  if (encoding === "pcm16") return int16ToFloat(view.getInt16(offset, true));
+  if (encoding === "pcmf32") return finite(view.getFloat32(offset, true));
+  if (encoding === "mulaw") return int16ToFloat(mulawToLinear(view.getUint8(offset)));
+  return int16ToFloat(alawToLinear(view.getUint8(offset)));
+}
+function encode(channels, format) {
+  if (format.encoding === "wav") {
+    const pcm = encode(channels, { ...format, encoding: "pcm16" });
+    return wavBytes(pcm, format.sampleRate, format.channels);
+  }
+  const width = sampleWidth(format.encoding);
+  const frames = channels[0]?.length ?? 0;
+  const bytes = new Uint8Array(frames * format.channels * width);
+  const view = new DataView(bytes.buffer);
+  for (let frame = 0; frame < frames; frame++) {
+    for (let channel = 0; channel < format.channels; channel++) {
+      const offset = (frame * format.channels + channel) * width;
+      writeSample(view, offset, format.encoding, channels[channel]?.[frame] ?? 0);
+    }
+  }
+  return bytes;
+}
+function writeSample(view, offset, encoding, sample) {
+  const unit = finite(sample);
+  if (encoding === "pcm16") view.setInt16(offset, floatToInt16(unit), true);
+  else if (encoding === "pcmf32") view.setFloat32(offset, unit, true);
+  else if (encoding === "mulaw") view.setUint8(offset, linearToMulaw(floatToInt16(unit)));
+  else view.setUint8(offset, linearToAlaw(floatToInt16(unit)));
+}
+function mix(channels, target) {
+  const length = channels[0]?.length ?? 0;
+  if (channels.length === target) return channels;
+  if (target === 1) {
+    const mixed = new Float32Array(length);
+    if (channels.length === 0) return [mixed];
+    for (let index = 0; index < length; index++) {
+      let sum = 0;
+      for (const channel of channels) sum += channel[index] ?? 0;
+      mixed[index] = sum / channels.length;
+    }
+    return [mixed];
+  }
+  const left = channels[0] ?? new Float32Array(length);
+  return [left, channels[1] ?? left];
+}
+function resample(input, inRate, outRate) {
+  if (inRate === outRate || input.length === 0) return input;
+  const outLength = Math.max(1, Math.round(input.length * outRate / inRate));
+  const output = new Float32Array(outLength);
+  const scale = inRate / outRate;
+  const last = input.length - 1;
+  for (let index = 0; index < outLength; index++) {
+    const position = index * scale;
+    const cursor = Math.min(Math.floor(position), last);
+    const next = Math.min(cursor + 1, last);
+    const fraction = position - cursor;
+    const left = input[cursor] ?? 0;
+    const right = input[next] ?? 0;
+    output[index] = left + (right - left) * fraction;
+  }
+  return output;
+}
+function finite(sample) {
+  return Number.isFinite(sample) ? sample : 0;
+}
+function floatToInt16(sample) {
+  const clamped = Math.max(-1, Math.min(1, sample));
+  const scaled = clamped < 0 ? clamped * 32768 : clamped * 32767;
+  return Math.max(-32768, Math.min(32767, Math.round(scaled)));
+}
+function int16ToFloat(sample) {
+  return sample < 0 ? sample / 32768 : sample / 32767;
+}
+function mulawToLinear(mulaw) {
+  const value = ~mulaw;
+  const exponent = (value & 112) >> 4;
+  let sample = ((value & 15) << 3) + LAW_BIAS;
+  sample <<= exponent;
+  return (value & 128) !== 0 ? LAW_BIAS - sample : sample - LAW_BIAS;
+}
+function alawToLinear(alaw) {
+  const value = alaw ^ 85;
+  const exponent = (value & 112) >> 4;
+  const mantissa = value & 15;
+  const sample = exponent === 0 ? mantissa + mantissa + 1 << 3 : mantissa + mantissa + 1 + 32 << exponent + 2;
+  return (value & 128) !== 0 ? sample : -sample;
+}
+function buildLawTable(expand, mask) {
+  const table = new Uint8Array(16384);
+  table[8192] = mask;
+  let cursor = 1;
+  for (let code = 0; code < 127; code++) {
+    const left = expand(code ^ mask);
+    const right = expand(code + 1 ^ mask);
+    const limit = left + right + 4 >> 3;
+    for (; cursor < limit; cursor += 1) {
+      table[8192 - cursor] = (code ^ (mask ^ 128)) & 255;
+      table[8192 + cursor] = (code ^ mask) & 255;
+    }
+  }
+  for (; cursor < 8192; cursor++) {
+    table[8192 - cursor] = (127 ^ (mask ^ 128)) & 255;
+    table[8192 + cursor] = (127 ^ mask) & 255;
+  }
+  table[0] = table[1] ?? mask;
+  return table;
+}
+const mulawTable = buildLawTable(mulawToLinear, 255);
+const alawTable = buildLawTable(alawToLinear, 213);
+function lawCode(table, pcm, fallback) {
+  return table[pcm + 32768 >> 2] ?? fallback;
+}
+function linearToMulaw(pcm) {
+  return lawCode(mulawTable, pcm, 255);
+}
+function linearToAlaw(pcm) {
+  return lawCode(alawTable, pcm, 213);
+}
+function decodeWav(data) {
+  if (data.byteLength < 44) throw new Error("Invalid WAV audio");
+  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);
+  if (text(data, 0, 4) !== "RIFF" || text(data, 8, 4) !== "WAVE")
+    throw new Error("Invalid WAV audio");
+  let cursor = 12;
+  let format;
+  let payload;
+  while (cursor + 8 <= data.byteLength) {
+    const id = text(data, cursor, 4);
+    const size = view.getUint32(cursor + 4, true);
+    const start = cursor + 8;
+    const end = Math.min(start + size, data.byteLength);
+    if (id === "fmt " && end - start >= 16) {
+      format = {
+        audioFormat: view.getUint16(start, true),
+        channels: view.getUint16(start + 2, true),
+        sampleRate: view.getUint32(start + 4, true),
+        width: view.getUint16(start + 14, true) / 8
+      };
+    } else if (id === "data") payload = data.subarray(start, end);
+    cursor = start + size + size % 2;
+  }
+  if (!format || !payload) throw new Error("Invalid WAV audio");
+  if (format.audioFormat !== 1 || format.width !== 2) throw new Error("WAV audio must be PCM16");
+  if (format.channels !== 1 && format.channels !== 2) throw new Error("Invalid WAV audio");
+  return {
+    channels: decodePcm(payload, {
+      encoding: "pcm16",
+      sampleRate: format.sampleRate,
+      channels: format.channels
+    }),
+    sampleRate: format.sampleRate
+  };
+}
+function wavBytes(pcm, sampleRate, channels) {
+  const bytes = new Uint8Array(44 + pcm.byteLength);
+  const view = new DataView(bytes.buffer);
+  bytes.set(ascii("RIFF"), 0);
+  view.setUint32(4, 36 + pcm.byteLength, true);
+  bytes.set(ascii("WAVE"), 8);
+  bytes.set(ascii("fmt "), 12);
+  view.setUint32(16, 16, true);
+  view.setUint16(20, 1, true);
+  view.setUint16(22, channels, true);
+  view.setUint32(24, sampleRate, true);
+  view.setUint32(28, sampleRate * channels * 2, true);
+  view.setUint16(32, channels * 2, true);
+  view.setUint16(34, 16, true);
+  bytes.set(ascii("data"), 36);
+  view.setUint32(40, pcm.byteLength, true);
+  bytes.set(pcm, 44);
+  return bytes;
+}
+function text(data, offset, length) {
+  return String.fromCharCode(...data.subarray(offset, offset + length));
+}
+function ascii(value) {
+  return Uint8Array.from(value, (character) => character.charCodeAt(0));
+}
+export {
+  convertAudio,
+  convertAudioSource,
+  decodeAudioBase64,
+  encodeAudioBase64
+};
+//# sourceMappingURL=convert.js.map

package/dist/convert.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/convert.ts"],
+  "sourcesContent": ["import type { AudioChunk, AudioEncoding, AudioFormat, AudioSource } from './audio.ts';\n\nconst compressed = new Set<AudioEncoding>(['opus', 'mp3', 'flac']);\nconst LAW_BIAS = 0x84;\n\nconst base64Alphabet = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/';\n\nfunction byteAt(data: Uint8Array, index: number): number {\n  const byte = data[index];\n  return byte === undefined ? 0 : byte;\n}\n\nfunction symbolAt(index: number): string {\n  return base64Alphabet.charAt(index);\n}\n\nexport function encodeAudioBase64(data: Uint8Array): string {\n  let encoded = '';\n  for (let index = 0; index < data.length; index += 3) {\n    const second = data[index + 1];\n    const third = data[index + 2];\n    const triple =\n      (byteAt(data, index) << 16) | (byteAt(data, index + 1) << 8) | byteAt(data, index + 2);\n    encoded += symbolAt((triple >> 18) & 63);\n    encoded += symbolAt((triple >> 12) & 63);\n    encoded += second === undefined ? '=' : symbolAt((triple >> 6) & 63);\n    encoded += third === undefined ? '=' : symbolAt(triple & 63);\n  }\n  return encoded;\n}\n\nfunction codeAt(value: string, index: number): number {\n  const character = value[index];\n  if (character === undefined) return -1;\n  const code = base64Alphabet.indexOf(character);\n  if (code < 0) throw new Error('Invalid base64 audio');\n  return code;\n}\n\nfunction sextet(code: number): number {\n  return code < 0 ? 0 : code;\n}\n\nexport function decodeAudioBase64(value: string): Uint8Array {\n  const cleaned = value.replaceAll('=', '');\n  if (cleaned.length % 4 === 1) throw new Error('Invalid base64 audio');\n  const bytes = [];\n  for (let index = 0; index < cleaned.length; index += 4) {\n    const first = codeAt(cleaned, index);\n    const second = codeAt(cleaned, index + 1);\n    const third = codeAt(cleaned, index + 2);\n    const fourth = codeAt(cleaned, index + 3);\n    const triple =\n      (sextet(first) << 18) | (sextet(second) << 12) | (sextet(third) << 6) | sextet(fourth);\n    bytes.push((triple >> 16) & 0xff);\n    if (third >= 0) bytes.push((triple >> 8) & 0xff);\n    if (fourth >= 0) bytes.push(triple & 0xff);\n  }\n  return Uint8Array.from(bytes);\n}\n\nexport function convertAudio(chunk: AudioChunk, format: AudioFormat): AudioChunk {\n  checkFormat(chunk.format);\n  checkFormat(format);\n  if (sameFormat(chunk.format, format)) {\n    return { data: chunk.data.slice(), format, timestamp: chunk.timestamp };\n  }\n  if (compressed.has(chunk.format.encoding) || compressed.has(format.encoding)) {\n    throw new Error(\n      `Cannot convert ${chunk.format.encoding} to ${format.encoding} without a codec`\n    );\n  }\n  const decoded = decode(chunk.data, chunk.format);\n  const mixed = mix(decoded.channels, format.channels);\n  const resampled = mixed.map((channel) =>\n    resample(channel, decoded.sampleRate, format.sampleRate)\n  );\n  return {\n    data: encode(resampled, format),\n    format,\n    timestamp: chunk.timestamp,\n  };\n}\n\nexport async function* convertAudioSource(source: AudioSource, format: AudioFormat): AudioSource {\n  for await (const chunk of source) yield convertAudio(chunk, format);\n}\n\nfunction sameFormat(left: AudioFormat, right: AudioFormat): boolean {\n  return (\n    left.encoding === right.encoding &&\n    left.sampleRate === right.sampleRate &&\n    left.channels === right.channels\n  );\n}\n\nfunction checkFormat(format: AudioFormat): void {\n  if (format.channels !== 1 && format.channels !== 2) {\n    throw new Error(`Unsupported channel count ${String(format.channels)}`);\n  }\n  if (!Number.isFinite(format.sampleRate) || format.sampleRate <= 0) {\n    throw new Error(`Unsupported sample rate ${String(format.sampleRate)}`);\n  }\n}\n\ninterface Planar {\n  channels: Float32Array[];\n  sampleRate: number;\n}\n\nfunction decode(data: Uint8Array, format: AudioFormat): Planar {\n  if (format.encoding === 'wav') return decodeWav(data);\n  return { channels: decodePcm(data, format), sampleRate: format.sampleRate };\n}\n\nfunction decodePcm(data: Uint8Array, format: AudioFormat): Float32Array[] {\n  const width = sampleWidth(format.encoding);\n  const frames = Math.floor(data.byteLength / (width * format.channels));\n  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);\n  const channels = Array.from({ length: format.channels }, () => new Float32Array(frames));\n  for (let frame = 0; frame < frames; frame++) {\n    for (let channel = 0; channel < format.channels; channel++) {\n      const offset = (frame * format.channels + channel) * width;\n      channels[channel][frame] = readSample(view, offset, format.encoding);\n    }\n  }\n  return channels;\n}\n\nfunction sampleWidth(encoding: AudioEncoding): number {\n  if (encoding === 'pcm16') return 2;\n  if (encoding === 'pcmf32') return 4;\n  if (encoding === 'mulaw' || encoding === 'alaw') return 1;\n  throw new Error(`Unsupported audio encoding ${encoding}`);\n}\n\nfunction readSample(view: DataView, offset: number, encoding: AudioEncoding): number {\n  if (encoding === 'pcm16') return int16ToFloat(view.getInt16(offset, true));\n  if (encoding === 'pcmf32') return finite(view.getFloat32(offset, true));\n  if (encoding === 'mulaw') return int16ToFloat(mulawToLinear(view.getUint8(offset)));\n  return int16ToFloat(alawToLinear(view.getUint8(offset)));\n}\n\nfunction encode(channels: Float32Array[], format: AudioFormat): Uint8Array {\n  if (format.encoding === 'wav') {\n    const pcm = encode(channels, { ...format, encoding: 'pcm16' });\n    return wavBytes(pcm, format.sampleRate, format.channels);\n  }\n  const width = sampleWidth(format.encoding);\n  const frames = channels[0]?.length ?? 0;\n  const bytes = new Uint8Array(frames * format.channels * width);\n  const view = new DataView(bytes.buffer);\n  for (let frame = 0; frame < frames; frame++) {\n    for (let channel = 0; channel < format.channels; channel++) {\n      const offset = (frame * format.channels + channel) * width;\n      writeSample(view, offset, format.encoding, channels[channel]?.[frame] ?? 0);\n    }\n  }\n  return bytes;\n}\n\nfunction writeSample(\n  view: DataView,\n  offset: number,\n  encoding: AudioEncoding,\n  sample: number\n): void {\n  const unit = finite(sample);\n  if (encoding === 'pcm16') view.setInt16(offset, floatToInt16(unit), true);\n  else if (encoding === 'pcmf32') view.setFloat32(offset, unit, true);\n  else if (encoding === 'mulaw') view.setUint8(offset, linearToMulaw(floatToInt16(unit)));\n  else view.setUint8(offset, linearToAlaw(floatToInt16(unit)));\n}\n\nfunction mix(channels: Float32Array[], target: 1 | 2): Float32Array[] {\n  const length = channels[0]?.length ?? 0;\n  if (channels.length === target) return channels;\n  if (target === 1) {\n    const mixed = new Float32Array(length);\n    if (channels.length === 0) return [mixed];\n    for (let index = 0; index < length; index++) {\n      let sum = 0;\n      for (const channel of channels) sum += channel[index] ?? 0;\n      mixed[index] = sum / channels.length;\n    }\n    return [mixed];\n  }\n  const left = channels[0] ?? new Float32Array(length);\n  return [left, channels[1] ?? left];\n}\n\nfunction resample(input: Float32Array, inRate: number, outRate: number): Float32Array {\n  if (inRate === outRate || input.length === 0) return input;\n  const outLength = Math.max(1, Math.round((input.length * outRate) / inRate));\n  const output = new Float32Array(outLength);\n  const scale = inRate / outRate;\n  const last = input.length - 1;\n  for (let index = 0; index < outLength; index++) {\n    const position = index * scale;\n    const cursor = Math.min(Math.floor(position), last);\n    const next = Math.min(cursor + 1, last);\n    const fraction = position - cursor;\n    const left = input[cursor] ?? 0;\n    const right = input[next] ?? 0;\n    output[index] = left + (right - left) * fraction;\n  }\n  return output;\n}\n\nfunction finite(sample: number): number {\n  return Number.isFinite(sample) ? sample : 0;\n}\n\nfunction floatToInt16(sample: number): number {\n  const clamped = Math.max(-1, Math.min(1, sample));\n  const scaled = clamped < 0 ? clamped * 0x8000 : clamped * 0x7fff;\n  return Math.max(-0x8000, Math.min(0x7fff, Math.round(scaled)));\n}\n\nfunction int16ToFloat(sample: number): number {\n  return sample < 0 ? sample / 0x8000 : sample / 0x7fff;\n}\n\nfunction mulawToLinear(mulaw: number): number {\n  const value = ~mulaw;\n  const exponent = (value & 0x70) >> 4;\n  let sample = ((value & 0x0f) << 3) + LAW_BIAS;\n  sample <<= exponent;\n  return (value & 0x80) !== 0 ? LAW_BIAS - sample : sample - LAW_BIAS;\n}\n\nfunction alawToLinear(alaw: number): number {\n  const value = alaw ^ 0x55;\n  const exponent = (value & 0x70) >> 4;\n  const mantissa = value & 0x0f;\n  const sample =\n    exponent === 0\n      ? (mantissa + mantissa + 1) << 3\n      : (mantissa + mantissa + 1 + 32) << (exponent + 2);\n  return (value & 0x80) !== 0 ? sample : -sample;\n}\n\nfunction buildLawTable(expand: (code: number) => number, mask: number): Uint8Array {\n  const table = new Uint8Array(16384);\n  table[8192] = mask;\n  let cursor = 1;\n  for (let code = 0; code < 127; code++) {\n    const left = expand(code ^ mask);\n    const right = expand((code + 1) ^ mask);\n    const limit = (left + right + 4) >> 3;\n    for (; cursor < limit; cursor += 1) {\n      table[8192 - cursor] = (code ^ (mask ^ 0x80)) & 0xff;\n      table[8192 + cursor] = (code ^ mask) & 0xff;\n    }\n  }\n  for (; cursor < 8192; cursor++) {\n    table[8192 - cursor] = (127 ^ (mask ^ 0x80)) & 0xff;\n    table[8192 + cursor] = (127 ^ mask) & 0xff;\n  }\n  table[0] = table[1] ?? mask;\n  return table;\n}\n\nconst mulawTable = buildLawTable(mulawToLinear, 0xff);\nconst alawTable = buildLawTable(alawToLinear, 0xd5);\n\nfunction lawCode(table: Uint8Array, pcm: number, fallback: number): number {\n  return table[(pcm + 32768) >> 2] ?? fallback;\n}\n\nfunction linearToMulaw(pcm: number): number {\n  return lawCode(mulawTable, pcm, 0xff);\n}\n\nfunction linearToAlaw(pcm: number): number {\n  return lawCode(alawTable, pcm, 0xd5);\n}\n\nfunction decodeWav(data: Uint8Array): Planar {\n  if (data.byteLength < 44) throw new Error('Invalid WAV audio');\n  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);\n  if (text(data, 0, 4) !== 'RIFF' || text(data, 8, 4) !== 'WAVE')\n    throw new Error('Invalid WAV audio');\n  let cursor = 12;\n  let format:\n    | { audioFormat: number; channels: number; sampleRate: number; width: number }\n    | undefined;\n  let payload: Uint8Array | undefined;\n  while (cursor + 8 <= data.byteLength) {\n    const id = text(data, cursor, 4);\n    const size = view.getUint32(cursor + 4, true);\n    const start = cursor + 8;\n    const end = Math.min(start + size, data.byteLength);\n    if (id === 'fmt ' && end - start >= 16) {\n      format = {\n        audioFormat: view.getUint16(start, true),\n        channels: view.getUint16(start + 2, true),\n        sampleRate: view.getUint32(start + 4, true),\n        width: view.getUint16(start + 14, true) / 8,\n      };\n    } else if (id === 'data') payload = data.subarray(start, end);\n    cursor = start + size + (size % 2);\n  }\n  if (!format || !payload) throw new Error('Invalid WAV audio');\n  if (format.audioFormat !== 1 || format.width !== 2) throw new Error('WAV audio must be PCM16');\n  if (format.channels !== 1 && format.channels !== 2) throw new Error('Invalid WAV audio');\n  return {\n    channels: decodePcm(payload, {\n      encoding: 'pcm16',\n      sampleRate: format.sampleRate,\n      channels: format.channels,\n    }),\n    sampleRate: format.sampleRate,\n  };\n}\n\nfunction wavBytes(pcm: Uint8Array, sampleRate: number, channels: number): Uint8Array {\n  const bytes = new Uint8Array(44 + pcm.byteLength);\n  const view = new DataView(bytes.buffer);\n  bytes.set(ascii('RIFF'), 0);\n  view.setUint32(4, 36 + pcm.byteLength, true);\n  bytes.set(ascii('WAVE'), 8);\n  bytes.set(ascii('fmt '), 12);\n  view.setUint32(16, 16, true);\n  view.setUint16(20, 1, true);\n  view.setUint16(22, channels, true);\n  view.setUint32(24, sampleRate, true);\n  view.setUint32(28, sampleRate * channels * 2, true);\n  view.setUint16(32, channels * 2, true);\n  view.setUint16(34, 16, true);\n  bytes.set(ascii('data'), 36);\n  view.setUint32(40, pcm.byteLength, true);\n  bytes.set(pcm, 44);\n  return bytes;\n}\n\nfunction text(data: Uint8Array, offset: number, length: number): string {\n  return String.fromCharCode(...data.subarray(offset, offset + length));\n}\n\nfunction ascii(value: string): Uint8Array {\n  return Uint8Array.from(value, (character) => character.charCodeAt(0));\n}\n"],
+  "mappings": "AAEA,MAAM,aAAa,oBAAI,IAAmB,CAAC,QAAQ,OAAO,MAAM,CAAC;AACjE,MAAM,WAAW;AAEjB,MAAM,iBAAiB;AAEvB,SAAS,OAAO,MAAkB,OAAuB;AACvD,QAAM,OAAO,KAAK,KAAK;AACvB,SAAO,SAAS,SAAY,IAAI;AAClC;AAEA,SAAS,SAAS,OAAuB;AACvC,SAAO,eAAe,OAAO,KAAK;AACpC;AAEO,SAAS,kBAAkB,MAA0B;AAC1D,MAAI,UAAU;AACd,WAAS,QAAQ,GAAG,QAAQ,KAAK,QAAQ,SAAS,GAAG;AACnD,UAAM,SAAS,KAAK,QAAQ,CAAC;AAC7B,UAAM,QAAQ,KAAK,QAAQ,CAAC;AAC5B,UAAM,SACH,OAAO,MAAM,KAAK,KAAK,KAAO,OAAO,MAAM,QAAQ,CAAC,KAAK,IAAK,OAAO,MAAM,QAAQ,CAAC;AACvF,eAAW,SAAU,UAAU,KAAM,EAAE;AACvC,eAAW,SAAU,UAAU,KAAM,EAAE;AACvC,eAAW,WAAW,SAAY,MAAM,SAAU,UAAU,IAAK,EAAE;AACnE,eAAW,UAAU,SAAY,MAAM,SAAS,SAAS,EAAE;AAAA,EAC7D;AACA,SAAO;AACT;AAEA,SAAS,OAAO,OAAe,OAAuB;AACpD,QAAM,YAAY,MAAM,KAAK;AAC7B,MAAI,cAAc,OAAW,QAAO;AACpC,QAAM,OAAO,eAAe,QAAQ,SAAS;AAC7C,MAAI,OAAO,EAAG,OAAM,IAAI,MAAM,sBAAsB;AACpD,SAAO;AACT;AAEA,SAAS,OAAO,MAAsB;AACpC,SAAO,OAAO,IAAI,IAAI;AACxB;AAEO,SAAS,kBAAkB,OAA2B;AAC3D,QAAM,UAAU,MAAM,WAAW,KAAK,EAAE;AACxC,MAAI,QAAQ,SAAS,MAAM,EAAG,OAAM,IAAI,MAAM,sBAAsB;AACpE,QAAM,QAAQ,CAAC;AACf,WAAS,QAAQ,GAAG,QAAQ,QAAQ,QAAQ,SAAS,GAAG;AACtD,UAAM,QAAQ,OAAO,SAAS,KAAK;AACnC,UAAM,SAAS,OAAO,SAAS,QAAQ,CAAC;AACxC,UAAM,QAAQ,OAAO,SAAS,QAAQ,CAAC;AACvC,UAAM,SAAS,OAAO,SAAS,QAAQ,CAAC;AACxC,UAAM,SACH,OAAO,KAAK,KAAK,KAAO,OAAO,MAAM,KAAK,KAAO,OAAO,KAAK,KAAK,IAAK,OAAO,MAAM;AACvF,UAAM,KAAM,UAAU,KAAM,GAAI;AAChC,QAAI,SAAS,EAAG,OAAM,KAAM,UAAU,IAAK,GAAI;AAC/C,QAAI,UAAU,EAAG,OAAM,KAAK,SAAS,GAAI;AAAA,EAC3C;AACA,SAAO,WAAW,KAAK,KAAK;AAC9B;AAEO,SAAS,aAAa,OAAmB,QAAiC;AAC/E,cAAY,MAAM,MAAM;AACxB,cAAY,MAAM;AAClB,MAAI,WAAW,MAAM,QAAQ,MAAM,GAAG;AACpC,WAAO,EAAE,MAAM,MAAM,KAAK,MAAM,GAAG,QAAQ,WAAW,MAAM,UAAU;AAAA,EACxE;AACA,MAAI,WAAW,IAAI,MAAM,OAAO,QAAQ,KAAK,WAAW,IAAI,OAAO,QAAQ,GAAG;AAC5E,UAAM,IAAI;AAAA,MACR,kBAAkB,MAAM,OAAO,QAAQ,OAAO,OAAO,QAAQ;AAAA,IAC/D;AAAA,EACF;AACA,QAAM,UAAU,OAAO,MAAM,MAAM,MAAM,MAAM;AAC/C,QAAM,QAAQ,IAAI,QAAQ,UAAU,OAAO,QAAQ;AACnD,QAAM,YAAY,MAAM;AAAA,IAAI,CAAC,YAC3B,SAAS,SAAS,QAAQ,YAAY,OAAO,UAAU;AAAA,EACzD;AACA,SAAO;AAAA,IACL,MAAM,OAAO,WAAW,MAAM;AAAA,IAC9B;AAAA,IACA,WAAW,MAAM;AAAA,EACnB;AACF;AAEA,gBAAuB,mBAAmB,QAAqB,QAAkC;AAC/F,mBAAiB,SAAS,OAAQ,OAAM,aAAa,OAAO,MAAM;AACpE;AAEA,SAAS,WAAW,MAAmB,OAA6B;AAClE,SACE,KAAK,aAAa,MAAM,YACxB,KAAK,eAAe,MAAM,cAC1B,KAAK,aAAa,MAAM;AAE5B;AAEA,SAAS,YAAY,QAA2B;AAC9C,MAAI,OAAO,aAAa,KAAK,OAAO,aAAa,GAAG;AAClD,UAAM,IAAI,MAAM,6BAA6B,OAAO,OAAO,QAAQ,CAAC,EAAE;AAAA,EACxE;AACA,MAAI,CAAC,OAAO,SAAS,OAAO,UAAU,KAAK,OAAO,cAAc,GAAG;AACjE,UAAM,IAAI,MAAM,2BAA2B,OAAO,OAAO,UAAU,CAAC,EAAE;AAAA,EACxE;AACF;AAOA,SAAS,OAAO,MAAkB,QAA6B;AAC7D,MAAI,OAAO,aAAa,MAAO,QAAO,UAAU,IAAI;AACpD,SAAO,EAAE,UAAU,UAAU,MAAM,MAAM,GAAG,YAAY,OAAO,WAAW;AAC5E;AAEA,SAAS,UAAU,MAAkB,QAAqC;AACxE,QAAM,QAAQ,YAAY,OAAO,QAAQ;AACzC,QAAM,SAAS,KAAK,MAAM,KAAK,cAAc,QAAQ,OAAO,SAAS;AACrE,QAAM,OAAO,IAAI,SAAS,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU;AACvE,QAAM,WAAW,MAAM,KAAK,EAAE,QAAQ,OAAO,SAAS,GAAG,MAAM,IAAI,aAAa,MAAM,CAAC;AACvF,WAAS,QAAQ,GAAG,QAAQ,QAAQ,SAAS;AAC3C,aAAS,UAAU,GAAG,UAAU,OAAO,UAAU,WAAW;AAC1D,YAAM,UAAU,QAAQ,OAAO,WAAW,WAAW;AACrD,eAAS,OAAO,EAAE,KAAK,IAAI,WAAW,MAAM,QAAQ,OAAO,QAAQ;AAAA,IACrE;AAAA,EACF;AACA,SAAO;AACT;AAEA,SAAS,YAAY,UAAiC;AACpD,MAAI,aAAa,QAAS,QAAO;AACjC,MAAI,aAAa,SAAU,QAAO;AAClC,MAAI,aAAa,WAAW,aAAa,OAAQ,QAAO;AACxD,QAAM,IAAI,MAAM,8BAA8B,QAAQ,EAAE;AAC1D;AAEA,SAAS,WAAW,MAAgB,QAAgB,UAAiC;AACnF,MAAI,aAAa,QAAS,QAAO,aAAa,KAAK,SAAS,QAAQ,IAAI,CAAC;AACzE,MAAI,aAAa,SAAU,QAAO,OAAO,KAAK,WAAW,QAAQ,IAAI,CAAC;AACtE,MAAI,aAAa,QAAS,QAAO,aAAa,cAAc,KAAK,SAAS,MAAM,CAAC,CAAC;AAClF,SAAO,aAAa,aAAa,KAAK,SAAS,MAAM,CAAC,CAAC;AACzD;AAEA,SAAS,OAAO,UAA0B,QAAiC;AACzE,MAAI,OAAO,aAAa,OAAO;AAC7B,UAAM,MAAM,OAAO,UAAU,EAAE,GAAG,QAAQ,UAAU,QAAQ,CAAC;AAC7D,WAAO,SAAS,KAAK,OAAO,YAAY,OAAO,QAAQ;AAAA,EACzD;AACA,QAAM,QAAQ,YAAY,OAAO,QAAQ;AACzC,QAAM,SAAS,SAAS,CAAC,GAAG,UAAU;AACtC,QAAM,QAAQ,IAAI,WAAW,SAAS,OAAO,WAAW,KAAK;AAC7D,QAAM,OAAO,IAAI,SAAS,MAAM,MAAM;AACtC,WAAS,QAAQ,GAAG,QAAQ,QAAQ,SAAS;AAC3C,aAAS,UAAU,GAAG,UAAU,OAAO,UAAU,WAAW;AAC1D,YAAM,UAAU,QAAQ,OAAO,WAAW,WAAW;AACrD,kBAAY,MAAM,QAAQ,OAAO,UAAU,SAAS,OAAO,IAAI,KAAK,KAAK,CAAC;AAAA,IAC5E;AAAA,EACF;AACA,SAAO;AACT;AAEA,SAAS,YACP,MACA,QACA,UACA,QACM;AACN,QAAM,OAAO,OAAO,MAAM;AAC1B,MAAI,aAAa,QAAS,MAAK,SAAS,QAAQ,aAAa,IAAI,GAAG,IAAI;AAAA,WAC/D,aAAa,SAAU,MAAK,WAAW,QAAQ,MAAM,IAAI;AAAA,WACzD,aAAa,QAAS,MAAK,SAAS,QAAQ,cAAc,aAAa,IAAI,CAAC,CAAC;AAAA,MACjF,MAAK,SAAS,QAAQ,aAAa,aAAa,IAAI,CAAC,CAAC;AAC7D;AAEA,SAAS,IAAI,UAA0B,QAA+B;AACpE,QAAM,SAAS,SAAS,CAAC,GAAG,UAAU;AACtC,MAAI,SAAS,WAAW,OAAQ,QAAO;AACvC,MAAI,WAAW,GAAG;AAChB,UAAM,QAAQ,IAAI,aAAa,MAAM;AACrC,QAAI,SAAS,WAAW,EAAG,QAAO,CAAC,KAAK;AACxC,aAAS,QAAQ,GAAG,QAAQ,QAAQ,SAAS;AAC3C,UAAI,MAAM;AACV,iBAAW,WAAW,SAAU,QAAO,QAAQ,KAAK,KAAK;AACzD,YAAM,KAAK,IAAI,MAAM,SAAS;AAAA,IAChC;AACA,WAAO,CAAC,KAAK;AAAA,EACf;AACA,QAAM,OAAO,SAAS,CAAC,KAAK,IAAI,aAAa,MAAM;AACnD,SAAO,CAAC,MAAM,SAAS,CAAC,KAAK,IAAI;AACnC;AAEA,SAAS,SAAS,OAAqB,QAAgB,SAA+B;AACpF,MAAI,WAAW,WAAW,MAAM,WAAW,EAAG,QAAO;AACrD,QAAM,YAAY,KAAK,IAAI,GAAG,KAAK,MAAO,MAAM,SAAS,UAAW,MAAM,CAAC;AAC3E,QAAM,SAAS,IAAI,aAAa,SAAS;AACzC,QAAM,QAAQ,SAAS;AACvB,QAAM,OAAO,MAAM,SAAS;AAC5B,WAAS,QAAQ,GAAG,QAAQ,WAAW,SAAS;AAC9C,UAAM,WAAW,QAAQ;AACzB,UAAM,SAAS,KAAK,IAAI,KAAK,MAAM,QAAQ,GAAG,IAAI;AAClD,UAAM,OAAO,KAAK,IAAI,SAAS,GAAG,IAAI;AACtC,UAAM,WAAW,WAAW;AAC5B,UAAM,OAAO,MAAM,MAAM,KAAK;AAC9B,UAAM,QAAQ,MAAM,IAAI,KAAK;AAC7B,WAAO,KAAK,IAAI,QAAQ,QAAQ,QAAQ;AAAA,EAC1C;AACA,SAAO;AACT;AAEA,SAAS,OAAO,QAAwB;AACtC,SAAO,OAAO,SAAS,MAAM,IAAI,SAAS;AAC5C;AAEA,SAAS,aAAa,QAAwB;AAC5C,QAAM,UAAU,KAAK,IAAI,IAAI,KAAK,IAAI,GAAG,MAAM,CAAC;AAChD,QAAM,SAAS,UAAU,IAAI,UAAU,QAAS,UAAU;AAC1D,SAAO,KAAK,IAAI,QAAS,KAAK,IAAI,OAAQ,KAAK,MAAM,MAAM,CAAC,CAAC;AAC/D;AAEA,SAAS,aAAa,QAAwB;AAC5C,SAAO,SAAS,IAAI,SAAS,QAAS,SAAS;AACjD;AAEA,SAAS,cAAc,OAAuB;AAC5C,QAAM,QAAQ,CAAC;AACf,QAAM,YAAY,QAAQ,QAAS;AACnC,MAAI,WAAW,QAAQ,OAAS,KAAK;AACrC,aAAW;AACX,UAAQ,QAAQ,SAAU,IAAI,WAAW,SAAS,SAAS;AAC7D;AAEA,SAAS,aAAa,MAAsB;AAC1C,QAAM,QAAQ,OAAO;AACrB,QAAM,YAAY,QAAQ,QAAS;AACnC,QAAM,WAAW,QAAQ;AACzB,QAAM,SACJ,aAAa,IACR,WAAW,WAAW,KAAM,IAC5B,WAAW,WAAW,IAAI,MAAQ,WAAW;AACpD,UAAQ,QAAQ,SAAU,IAAI,SAAS,CAAC;AAC1C;AAEA,SAAS,cAAc,QAAkC,MAA0B;AACjF,QAAM,QAAQ,IAAI,WAAW,KAAK;AAClC,QAAM,IAAI,IAAI;AACd,MAAI,SAAS;AACb,WAAS,OAAO,GAAG,OAAO,KAAK,QAAQ;AACrC,UAAM,OAAO,OAAO,OAAO,IAAI;AAC/B,UAAM,QAAQ,OAAQ,OAAO,IAAK,IAAI;AACtC,UAAM,QAAS,OAAO,QAAQ,KAAM;AACpC,WAAO,SAAS,OAAO,UAAU,GAAG;AAClC,YAAM,OAAO,MAAM,KAAK,QAAQ,OAAO,QAAS;AAChD,YAAM,OAAO,MAAM,KAAK,OAAO,QAAQ;AAAA,IACzC;AAAA,EACF;AACA,SAAO,SAAS,MAAM,UAAU;AAC9B,UAAM,OAAO,MAAM,KAAK,OAAO,OAAO,QAAS;AAC/C,UAAM,OAAO,MAAM,KAAK,MAAM,QAAQ;AAAA,EACxC;AACA,QAAM,CAAC,IAAI,MAAM,CAAC,KAAK;AACvB,SAAO;AACT;AAEA,MAAM,aAAa,cAAc,eAAe,GAAI;AACpD,MAAM,YAAY,cAAc,cAAc,GAAI;AAElD,SAAS,QAAQ,OAAmB,KAAa,UAA0B;AACzE,SAAO,MAAO,MAAM,SAAU,CAAC,KAAK;AACtC;AAEA,SAAS,cAAc,KAAqB;AAC1C,SAAO,QAAQ,YAAY,KAAK,GAAI;AACtC;AAEA,SAAS,aAAa,KAAqB;AACzC,SAAO,QAAQ,WAAW,KAAK,GAAI;AACrC;AAEA,SAAS,UAAU,MAA0B;AAC3C,MAAI,KAAK,aAAa,GAAI,OAAM,IAAI,MAAM,mBAAmB;AAC7D,QAAM,OAAO,IAAI,SAAS,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU;AACvE,MAAI,KAAK,MAAM,GAAG,CAAC,MAAM,UAAU,KAAK,MAAM,GAAG,CAAC,MAAM;AACtD,UAAM,IAAI,MAAM,mBAAmB;AACrC,MAAI,SAAS;AACb,MAAI;AAGJ,MAAI;AACJ,SAAO,SAAS,KAAK,KAAK,YAAY;AACpC,UAAM,KAAK,KAAK,MAAM,QAAQ,CAAC;AAC/B,UAAM,OAAO,KAAK,UAAU,SAAS,GAAG,IAAI;AAC5C,UAAM,QAAQ,SAAS;AACvB,UAAM,MAAM,KAAK,IAAI,QAAQ,MAAM,KAAK,UAAU;AAClD,QAAI,OAAO,UAAU,MAAM,SAAS,IAAI;AACtC,eAAS;AAAA,QACP,aAAa,KAAK,UAAU,OAAO,IAAI;AAAA,QACvC,UAAU,KAAK,UAAU,QAAQ,GAAG,IAAI;AAAA,QACxC,YAAY,KAAK,UAAU,QAAQ,GAAG,IAAI;AAAA,QAC1C,OAAO,KAAK,UAAU,QAAQ,IAAI,IAAI,IAAI;AAAA,MAC5C;AAAA,IACF,WAAW,OAAO,OAAQ,WAAU,KAAK,SAAS,OAAO,GAAG;AAC5D,aAAS,QAAQ,OAAQ,OAAO;AAAA,EAClC;AACA,MAAI,CAAC,UAAU,CAAC,QAAS,OAAM,IAAI,MAAM,mBAAmB;AAC5D,MAAI,OAAO,gBAAgB,KAAK,OAAO,UAAU,EAAG,OAAM,IAAI,MAAM,yBAAyB;AAC7F,MAAI,OAAO,aAAa,KAAK,OAAO,aAAa,EAAG,OAAM,IAAI,MAAM,mBAAmB;AACvF,SAAO;AAAA,IACL,UAAU,UAAU,SAAS;AAAA,MAC3B,UAAU;AAAA,MACV,YAAY,OAAO;AAAA,MACnB,UAAU,OAAO;AAAA,IACnB,CAAC;AAAA,IACD,YAAY,OAAO;AAAA,EACrB;AACF;AAEA,SAAS,SAAS,KAAiB,YAAoB,UAA8B;AACnF,QAAM,QAAQ,IAAI,WAAW,KAAK,IAAI,UAAU;AAChD,QAAM,OAAO,IAAI,SAAS,MAAM,MAAM;AACtC,QAAM,IAAI,MAAM,MAAM,GAAG,CAAC;AAC1B,OAAK,UAAU,GAAG,KAAK,IAAI,YAAY,IAAI;AAC3C,QAAM,IAAI,MAAM,MAAM,GAAG,CAAC;AAC1B,QAAM,IAAI,MAAM,MAAM,GAAG,EAAE;AAC3B,OAAK,UAAU,IAAI,IAAI,IAAI;AAC3B,OAAK,UAAU,IAAI,GAAG,IAAI;AAC1B,OAAK,UAAU,IAAI,UAAU,IAAI;AACjC,OAAK,UAAU,IAAI,YAAY,IAAI;AACnC,OAAK,UAAU,IAAI,aAAa,WAAW,GAAG,IAAI;AAClD,OAAK,UAAU,IAAI,WAAW,GAAG,IAAI;AACrC,OAAK,UAAU,IAAI,IAAI,IAAI;AAC3B,QAAM,IAAI,MAAM,MAAM,GAAG,EAAE;AAC3B,OAAK,UAAU,IAAI,IAAI,YAAY,IAAI;AACvC,QAAM,IAAI,KAAK,EAAE;AACjB,SAAO;AACT;AAEA,SAAS,KAAK,MAAkB,QAAgB,QAAwB;AACtE,SAAO,OAAO,aAAa,GAAG,KAAK,SAAS,QAAQ,SAAS,MAAM,CAAC;AACtE;AAEA,SAAS,MAAM,OAA2B;AACxC,SAAO,WAAW,KAAK,OAAO,CAAC,cAAc,UAAU,WAAW,CAAC,CAAC;AACtE;",
+  "names": []
+}

package/dist/index.js

@@ -1,5 +1,11 @@
 import { resolveVoiceAuth } from "./auth.js";
 import { blankUsage } from "./blank-usage.js";
+import {
+  convertAudio,
+  convertAudioSource,
+  decodeAudioBase64,
+  encodeAudioBase64
+} from "./convert.js";
 import { VoiceRequestError } from "./errors.js";
 import {
   getVoiceModel,
@@ -25,6 +31,10 @@ export {
   addAudioCounters,
   blankUsage,
   connectRealtime,
+  convertAudio,
+  convertAudioSource,
+  decodeAudioBase64,
+  encodeAudioBase64,
   getVoiceModel,
   listVoiceModels,
   listen,

package/dist/index.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/index.ts"],
-  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeToken,\n} from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\n"],
-  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AASP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;",
+  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeToken,\n} from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\n"],
+  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AASP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;",
   "names": []
 }

package/dist/types/convert.d.ts

@@ -0,0 +1,5 @@
+import type { AudioChunk, AudioFormat, AudioSource } from './audio.ts';
+export declare function encodeAudioBase64(data: Uint8Array): string;
+export declare function decodeAudioBase64(value: string): Uint8Array;
+export declare function convertAudio(chunk: AudioChunk, format: AudioFormat): AudioChunk;
+export declare function convertAudioSource(source: AudioSource, format: AudioFormat): AudioSource;

package/dist/types/index.d.ts

@@ -1,6 +1,7 @@
 export type { AudioChunk, AudioEncoding, AudioFormat, AudioSink, AudioSource, AudioTag, CharAlignment, TextSpan, } from './audio.ts';
 export { resolveVoiceAuth } from './auth.ts';
 export { blankUsage } from './blank-usage.ts';
+export { convertAudio, convertAudioSource, decodeAudioBase64, encodeAudioBase64, } from './convert.ts';
 export { VoiceRequestError } from './errors.ts';
 export type { Delivery, DeliverySupport, KnownRealtimeApi, KnownSpeechApi, KnownTranscriptionApi, RealtimeApi, RealtimeModel, SpeechApi, SpeechModel, TranscriptionApi, TranscriptionModel, VoiceInfo, VoiceModel, VoiceModelType, } from './models.ts';
 export { getVoiceModel, listVoiceModels, registerVoiceModel, } from './models.ts';

package/dist/types/worklet/convert.d.ts

@@ -0,0 +1,30 @@
+import type { AudioFormat } from '../audio.ts';
+export declare const AUDIO_CONVERT_PROCESSOR = "pi-audio-convert";
+export interface AudioConvertProcessorOptions {
+    processorOptions?: {
+        format?: AudioFormat;
+        sampleRate?: number;
+    };
+}
+interface WorkletScope {
+    AudioWorkletProcessor?: new (options?: AudioConvertProcessorOptions) => {
+        port?: {
+            postMessage(message: unknown): void;
+        };
+    };
+    registerProcessor?: (name: string, processor: new (options?: AudioConvertProcessorOptions) => object) => void;
+}
+declare const WorkletBase: new (options?: AudioConvertProcessorOptions) => {
+    port?: {
+        postMessage(message: unknown): void;
+    };
+};
+export declare class AudioConvertProcessor extends WorkletBase {
+    #private;
+    constructor(options?: AudioConvertProcessorOptions);
+    process(inputs: Array<ArrayLike<Float32Array | undefined> | undefined>): boolean;
+    take(): Uint8Array[];
+}
+export declare function registerAudioConvertWorklet(target?: WorkletScope): void;
+export declare function audioConvertWorkletUrl(): string;
+export {};

package/dist/worklet/convert.js

@@ -0,0 +1,80 @@
+import { convertAudio } from "../convert.js";
+const AUDIO_CONVERT_PROCESSOR = "pi-audio-convert";
+const scope = globalThis;
+const WorkletBase = scope.AudioWorkletProcessor ?? class {
+  port;
+};
+class AudioConvertProcessor extends WorkletBase {
+  #format;
+  #sampleRate;
+  #frames = 0;
+  #pending = [];
+  constructor(options) {
+    super(options);
+    this.#format = options?.processorOptions?.format ?? {
+      encoding: "pcm16",
+      sampleRate: 24e3,
+      channels: 1
+    };
+    const sampleRate = options?.processorOptions?.sampleRate;
+    this.#sampleRate = sampleRate !== void 0 && Number.isFinite(sampleRate) && sampleRate > 0 ? sampleRate : 48e3;
+  }
+  process(inputs) {
+    const input = inputs[0];
+    const channels = [];
+    if (input) {
+      for (let index = 0; index < input.length; index++) {
+        const channel = input[index];
+        if (channel) channels.push(channel);
+      }
+    }
+    const frames = channels[0]?.length ?? 0;
+    if (channels.length === 0 || frames === 0) return true;
+    const count = Math.min(channels.length, 2);
+    const data = new Uint8Array(frames * count * 4);
+    const view = new DataView(data.buffer);
+    for (let frame = 0; frame < frames; frame++) {
+      for (let channel = 0; channel < count; channel++) {
+        view.setFloat32((frame * count + channel) * 4, channels[channel]?.[frame] ?? 0, true);
+      }
+    }
+    const converted = convertAudio(
+      {
+        data,
+        format: {
+          encoding: "pcmf32",
+          sampleRate: this.#sampleRate,
+          channels: count === 1 ? 1 : 2
+        },
+        timestamp: this.#frames / this.#sampleRate * 1e3
+      },
+      this.#format
+    );
+    this.#frames += frames;
+    this.#pending.push(converted.data);
+    this.port?.postMessage({
+      data: converted.data,
+      format: converted.format,
+      timestamp: converted.timestamp
+    });
+    return true;
+  }
+  take() {
+    return this.#pending.splice(0);
+  }
+}
+function registerAudioConvertWorklet(target = scope) {
+  if (typeof target.registerProcessor !== "function") return;
+  target.registerProcessor(AUDIO_CONVERT_PROCESSOR, AudioConvertProcessor);
+}
+function audioConvertWorkletUrl() {
+  return new URL(import.meta.url).href;
+}
+registerAudioConvertWorklet();
+export {
+  AUDIO_CONVERT_PROCESSOR,
+  AudioConvertProcessor,
+  audioConvertWorkletUrl,
+  registerAudioConvertWorklet
+};
+//# sourceMappingURL=convert.js.map

package/dist/worklet/convert.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/worklet/convert.ts"],
+  "sourcesContent": ["import type { AudioFormat } from '../audio.ts';\nimport { convertAudio } from '../convert.ts';\n\nexport const AUDIO_CONVERT_PROCESSOR = 'pi-audio-convert';\n\nexport interface AudioConvertProcessorOptions {\n  processorOptions?: {\n    format?: AudioFormat;\n    sampleRate?: number;\n  };\n}\n\ninterface WorkletScope {\n  AudioWorkletProcessor?: new (\n    options?: AudioConvertProcessorOptions\n  ) => {\n    port?: { postMessage(message: unknown): void };\n  };\n  registerProcessor?: (\n    name: string,\n    processor: new (options?: AudioConvertProcessorOptions) => object\n  ) => void;\n}\n\nconst scope = globalThis as WorkletScope;\n\nconst WorkletBase =\n  scope.AudioWorkletProcessor ??\n  class {\n    port?: { postMessage(message: unknown): void };\n  };\n\nexport class AudioConvertProcessor extends WorkletBase {\n  readonly #format: AudioFormat;\n  readonly #sampleRate: number;\n  #frames = 0;\n  readonly #pending: Uint8Array[] = [];\n\n  constructor(options?: AudioConvertProcessorOptions) {\n    super(options);\n    this.#format = options?.processorOptions?.format ?? {\n      encoding: 'pcm16',\n      sampleRate: 24000,\n      channels: 1,\n    };\n    const sampleRate = options?.processorOptions?.sampleRate;\n    this.#sampleRate =\n      sampleRate !== undefined && Number.isFinite(sampleRate) && sampleRate > 0\n        ? sampleRate\n        : 48000;\n  }\n\n  process(inputs: Array<ArrayLike<Float32Array | undefined> | undefined>): boolean {\n    const input = inputs[0];\n    const channels: Float32Array[] = [];\n    if (input) {\n      for (let index = 0; index < input.length; index++) {\n        const channel = input[index];\n        if (channel) channels.push(channel);\n      }\n    }\n    const frames = channels[0]?.length ?? 0;\n    if (channels.length === 0 || frames === 0) return true;\n    const count = Math.min(channels.length, 2);\n    const data = new Uint8Array(frames * count * 4);\n    const view = new DataView(data.buffer);\n    for (let frame = 0; frame < frames; frame++) {\n      for (let channel = 0; channel < count; channel++) {\n        view.setFloat32((frame * count + channel) * 4, channels[channel]?.[frame] ?? 0, true);\n      }\n    }\n    const converted = convertAudio(\n      {\n        data,\n        format: {\n          encoding: 'pcmf32',\n          sampleRate: this.#sampleRate,\n          channels: count === 1 ? 1 : 2,\n        },\n        timestamp: (this.#frames / this.#sampleRate) * 1000,\n      },\n      this.#format\n    );\n    this.#frames += frames;\n    this.#pending.push(converted.data);\n    this.port?.postMessage({\n      data: converted.data,\n      format: converted.format,\n      timestamp: converted.timestamp,\n    });\n    return true;\n  }\n\n  take(): Uint8Array[] {\n    return this.#pending.splice(0);\n  }\n}\n\nexport function registerAudioConvertWorklet(target: WorkletScope = scope): void {\n  if (typeof target.registerProcessor !== 'function') return;\n  target.registerProcessor(AUDIO_CONVERT_PROCESSOR, AudioConvertProcessor);\n}\n\nexport function audioConvertWorkletUrl(): string {\n  return new URL(import.meta.url).href;\n}\n\nregisterAudioConvertWorklet();\n"],
+  "mappings": "AACA,SAAS,oBAAoB;AAEtB,MAAM,0BAA0B;AAqBvC,MAAM,QAAQ;AAEd,MAAM,cACJ,MAAM,yBACN,MAAM;AAAA,EACJ;AACF;AAEK,MAAM,8BAA8B,YAAY;AAAA,EAC5C;AAAA,EACA;AAAA,EACT,UAAU;AAAA,EACD,WAAyB,CAAC;AAAA,EAEnC,YAAY,SAAwC;AAClD,UAAM,OAAO;AACb,SAAK,UAAU,SAAS,kBAAkB,UAAU;AAAA,MAClD,UAAU;AAAA,MACV,YAAY;AAAA,MACZ,UAAU;AAAA,IACZ;AACA,UAAM,aAAa,SAAS,kBAAkB;AAC9C,SAAK,cACH,eAAe,UAAa,OAAO,SAAS,UAAU,KAAK,aAAa,IACpE,aACA;AAAA,EACR;AAAA,EAEA,QAAQ,QAAyE;AAC/E,UAAM,QAAQ,OAAO,CAAC;AACtB,UAAM,WAA2B,CAAC;AAClC,QAAI,OAAO;AACT,eAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS;AACjD,cAAM,UAAU,MAAM,KAAK;AAC3B,YAAI,QAAS,UAAS,KAAK,OAAO;AAAA,MACpC;AAAA,IACF;AACA,UAAM,SAAS,SAAS,CAAC,GAAG,UAAU;AACtC,QAAI,SAAS,WAAW,KAAK,WAAW,EAAG,QAAO;AAClD,UAAM,QAAQ,KAAK,IAAI,SAAS,QAAQ,CAAC;AACzC,UAAM,OAAO,IAAI,WAAW,SAAS,QAAQ,CAAC;AAC9C,UAAM,OAAO,IAAI,SAAS,KAAK,MAAM;AACrC,aAAS,QAAQ,GAAG,QAAQ,QAAQ,SAAS;AAC3C,eAAS,UAAU,GAAG,UAAU,OAAO,WAAW;AAChD,aAAK,YAAY,QAAQ,QAAQ,WAAW,GAAG,SAAS,OAAO,IAAI,KAAK,KAAK,GAAG,IAAI;AAAA,MACtF;AAAA,IACF;AACA,UAAM,YAAY;AAAA,MAChB;AAAA,QACE;AAAA,QACA,QAAQ;AAAA,UACN,UAAU;AAAA,UACV,YAAY,KAAK;AAAA,UACjB,UAAU,UAAU,IAAI,IAAI;AAAA,QAC9B;AAAA,QACA,WAAY,KAAK,UAAU,KAAK,cAAe;AAAA,MACjD;AAAA,MACA,KAAK;AAAA,IACP;AACA,SAAK,WAAW;AAChB,SAAK,SAAS,KAAK,UAAU,IAAI;AACjC,SAAK,MAAM,YAAY;AAAA,MACrB,MAAM,UAAU;AAAA,MAChB,QAAQ,UAAU;AAAA,MAClB,WAAW,UAAU;AAAA,IACvB,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAEA,OAAqB;AACnB,WAAO,KAAK,SAAS,OAAO,CAAC;AAAA,EAC/B;AACF;AAEO,SAAS,4BAA4B,SAAuB,OAAa;AAC9E,MAAI,OAAO,OAAO,sBAAsB,WAAY;AACpD,SAAO,kBAAkB,yBAAyB,qBAAqB;AACzE;AAEO,SAAS,yBAAiC;AAC/C,SAAO,IAAI,IAAI,YAAY,GAAG,EAAE;AAClC;AAEA,4BAA4B;",
+  "names": []
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.2.0",
+  "version": "1.3.0",
   "description": "Voice bindings for pi: speech-to-text, text-to-speech and realtime speech for pi-coding-agent plugins and slicc-agent",
   "type": "module",
   "license": "Apache-2.0",
@@ -29,6 +29,10 @@
     "./faux": {
       "types": "./dist/types/faux/index.d.ts",
       "default": "./dist/faux/index.js"
+    },
+    "./worklet": {
+      "types": "./dist/types/worklet/convert.d.ts",
+      "default": "./dist/worklet/convert.js"
     }
   },
   "files": [