use-voice-control 0.1.85 → 0.1.87

package/dist/cli.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.js","sources":["../speech/node/document.ts","../speech/node/render.ts","../speech/node/cli.ts"],"sourcesContent":["/**\n * @fileoverview Turning a document on disk (or on stdin) into speakable text.\n *\n * A `.md` file is converted with `markdownToSpeech` so the marks become\n * structure rather than words; a `.txt` file is passed through as it is. When\n * the source has no filename — piped stdin, a `--text` string — the format is\n * guessed from the content.\n */\nimport {\n  looksLikeMarkdown,\n  markdownToSpeech,\n  type MarkdownToSpeechOptions,\n} from \"../utils/markdown-to-speech\";\n\nexport type DocumentFormat = \"markdown\" | \"text\";\n/** What the caller asked for; `auto` defers to the extension, then the content. */\nexport type RequestedFormat = DocumentFormat | \"auto\";\n\n/** Extensions read as Markdown. Everything else is treated as plain text. */\nexport const MARKDOWN_EXTENSIONS = [\".md\", \".markdown\", \".mdown\", \".mkd\", \".mdx\"];\n\n/** Extensions the CLI accepts without complaint. Others still work, with a warning. */\nexport const TEXT_EXTENSIONS = [\".txt\", \".text\", \"\"];\n\n/** Lowercased extension of a path, including the dot (`\"\"` when there is none). */\nexport function extensionOf(filename: string): string {\n  const base = filename.split(/[\\\\/]/).pop() ?? \"\";\n  const dot = base.lastIndexOf(\".\");\n  return dot > 0 ? base.slice(dot).toLowerCase() : \"\";\n}\n\n/**\n * Decides how to read a document.\n *\n * An explicit `requested` format always wins — a `.txt` file full of Markdown is\n * still the user's call. Otherwise the extension decides, and content sniffing\n * is the last resort for input that arrived without a name.\n */\nexport function detectFormat(\n  content: string,\n  filename?: string,\n  requested: RequestedFormat = \"auto\"\n): DocumentFormat {\n  if (requested !== \"auto\") return requested;\n\n  if (filename) {\n    const extension = extensionOf(filename);\n    if (MARKDOWN_EXTENSIONS.includes(extension)) return \"markdown\";\n    if (TEXT_EXTENSIONS.includes(extension)) return \"text\";\n  }\n\n  return looksLikeMarkdown(content) ? \"markdown\" : \"text\";\n}\n\n/** Converts document content to the text that should be spoken. */\nexport function toSpeechText(\n  content: string,\n  format: DocumentFormat,\n  options: MarkdownToSpeechOptions = {}\n): string {\n  if (format === \"text\") {\n    // Plain text is already speakable; only normalise line endings and the\n    // trailing whitespace that would otherwise become an empty final chunk.\n    return content.replace(/\\r\\n?/g, \"\\n\").trim();\n  }\n  return markdownToSpeech(content, options);\n}\n\nexport interface LoadedDocument {\n  /** The speakable text, Markdown already converted. */\n  text: string;\n  /** How the content was read. */\n  format: DocumentFormat;\n  /** Where it came from, for log lines: a path, `stdin`, or `--text`. */\n  source: string;\n}\n\nexport interface LoadDocumentOptions {\n  /** Path to read. Use `-` for stdin. Mutually exclusive with `text`. */\n  file?: string;\n  /** Literal content to speak, instead of reading a file. */\n  text?: string;\n  format?: RequestedFormat;\n  markdown?: MarkdownToSpeechOptions;\n  /** Reads stdin. Injected so the CLI tests do not need a real pipe. */\n  readStdin?: () => Promise<string>;\n}\n\n/**\n * Reads a document from a path, stdin, or an inline string and returns the text\n * to synthesize.\n */\nexport async function loadDocument(\n  options: LoadDocumentOptions\n): Promise<LoadedDocument> {\n  const { file, text, format = \"auto\", markdown = {} } = options;\n\n  if (text !== undefined) {\n    const resolved = detectFormat(text, undefined, format);\n    return { text: toSpeechText(text, resolved, markdown), format: resolved, source: \"--text\" };\n  }\n\n  if (!file) throw new Error(\"No input: pass a file path, `-` for stdin, or --text\");\n\n  if (file === \"-\") {\n    const stdin = options.readStdin\n      ? await options.readStdin()\n      : await readStdinToString();\n    const resolved = detectFormat(stdin, undefined, format);\n    return { text: toSpeechText(stdin, resolved, markdown), format: resolved, source: \"stdin\" };\n  }\n\n  const { readFile } = await import(\"node:fs/promises\");\n  const content = await readFile(file, \"utf8\");\n  const resolved = detectFormat(content, file, format);\n  return { text: toSpeechText(content, resolved, markdown), format: resolved, source: file };\n}\n\n/** Collects all of stdin as UTF-8 text. */\nexport async function readStdinToString(): Promise<string> {\n  const chunks: Buffer[] = [];\n  for await (const chunk of process.stdin) {\n    chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));\n  }\n  return Buffer.concat(chunks).toString(\"utf8\");\n}\n","/**\n * @fileoverview Rendering a loaded document to an audio file on disk.\n *\n * Sits between `document.ts` (what to say) and `core/kokoro-node.ts` (how to say\n * it): resolves the output path, runs the synthesizer, and writes the WAV.\n */\nimport {\n  synthesizeSamples,\n  type KokoroAudio,\n  type KokoroNodeOptions,\n} from \"../core/kokoro-node\";\nimport { encodeWav, wavDurationSeconds } from \"../utils/wav\";\nimport { extensionOf, loadDocument, type LoadDocumentOptions } from \"./document\";\n\nexport interface RenderOptions extends LoadDocumentOptions, KokoroNodeOptions {\n  /** Where to write the audio. `-` writes the WAV to stdout. */\n  output?: string;\n  /**\n   * Replace the synthesizer. Used by the tests, and by hosts that already have a\n   * model loaded and do not want a second copy.\n   */\n  synthesize?: (text: string, options: KokoroNodeOptions) => Promise<KokoroAudio>;\n}\n\nexport interface RenderResult {\n  /** Path written, or `-` when the audio went to stdout. */\n  output: string;\n  /** The text that was spoken, after Markdown conversion. */\n  text: string;\n  format: \"markdown\" | \"text\";\n  source: string;\n  durationSeconds: number;\n  bytes: number;\n}\n\n/** `notes.md` becomes `notes.wav`, next to the original. */\nexport function defaultOutputPath(input: string): string {\n  const extension = extensionOf(input);\n  const base = extension ? input.slice(0, -extension.length) : input;\n  return `${base}.wav`;\n}\n\n/**\n * Reads a document, synthesizes it, and writes a WAV file.\n *\n * Returns the spoken text alongside the file details so a caller can show what\n * was read without converting the document a second time.\n */\nexport async function renderDocument(options: RenderOptions): Promise<RenderResult> {\n  const document = await loadDocument(options);\n\n  if (!document.text.trim()) {\n    throw new Error(`Nothing to speak: ${document.source} has no readable text`);\n  }\n\n  const synthesize = options.synthesize ?? synthesizeSamples;\n  const { samples, sampleRate } = await synthesize(document.text, options);\n  const audio = encodeWav(samples, sampleRate);\n\n  const output =\n    options.output ??\n    (options.file && options.file !== \"-\" ? defaultOutputPath(options.file) : \"out.wav\");\n\n  if (output === \"-\") {\n    process.stdout.write(Buffer.from(audio));\n  } else {\n    const { writeFile } = await import(\"node:fs/promises\");\n    await writeFile(output, Buffer.from(audio));\n  }\n\n  return {\n    output,\n    text: document.text,\n    format: document.format,\n    source: document.source,\n    durationSeconds: wavDurationSeconds(samples, sampleRate),\n    bytes: audio.byteLength,\n  };\n}\n","/**\n * @fileoverview The `use-voice-control` command line tool.\n *\n * Reads a Markdown or text file and speaks it to a WAV file:\n *\n * ```sh\n * npx use-voice-control README.md -o readme.wav\n * ```\n *\n * Argument parsing is a pure function (`parseArgs`) and every effect the command\n * has — writing lines, reading stdin, synthesizing — is injectable, so the whole\n * command is testable without loading a 90 MB model.\n */\nimport { describeKokoroVoice, KOKORO_VOICES } from \"../types/types\";\nimport type { KokoroDevice, KokoroDtype } from \"../core/kokoro-node\";\nimport type { MarkdownToSpeechOptions } from \"../utils/markdown-to-speech\";\nimport { loadDocument, type RequestedFormat } from \"./document\";\nimport { renderDocument, type RenderOptions } from \"./render\";\n\nexport type CliCommand = \"speak\" | \"print\" | \"help\" | \"version\" | \"list-voices\";\n\nexport interface CliOptions {\n  command: CliCommand;\n  /** Input path, or `-` for stdin. Undefined when `text` is set. */\n  input?: string;\n  /** Literal text passed with `--text`. */\n  text?: string;\n  output?: string;\n  format: RequestedFormat;\n  voice: string;\n  speed: number;\n  model?: string;\n  dtype?: KokoroDtype;\n  device?: KokoroDevice;\n  maxChunkLength?: number;\n  gapMs?: number;\n  markdown: MarkdownToSpeechOptions;\n  quiet: boolean;\n  /** Everything wrong with the command line. Non-empty means do not run. */\n  errors: string[];\n}\n\nexport interface CliIO {\n  /** Writes a line to stdout. */\n  log?: (line: string) => void;\n  /** Writes a line to stderr — progress and errors go here so `-o -` stays clean. */\n  error?: (line: string) => void;\n  /** Overrides synthesis, for tests. */\n  synthesize?: RenderOptions[\"synthesize\"];\n  /** Overrides reading stdin, for tests. */\n  readStdin?: () => Promise<string>;\n  /** Whether stdin is a terminal. When false and no input is given, stdin is read. */\n  stdinIsTTY?: boolean;\n}\n\nconst FORMATS: Record<string, RequestedFormat> = {\n  auto: \"auto\",\n  markdown: \"markdown\",\n  md: \"markdown\",\n  text: \"text\",\n  txt: \"text\",\n  plain: \"text\",\n};\n\nconst DTYPES: KokoroDtype[] = [\"fp32\", \"fp16\", \"q8\", \"q4\", \"q4f16\"];\nconst DEVICES: KokoroDevice[] = [\"wasm\", \"webgpu\", \"cpu\"];\n\nexport const USAGE = `use-voice-control — read a Markdown or text file aloud into an audio file\n\nUsage\n  npx use-voice-control <file.md|file.txt|-> [options]\n  npx use-voice-control --text \"Hello there\" -o hello.wav\n  cat notes.md | npx use-voice-control - -o notes.wav\n\nMarkdown is converted before it is spoken: \"#\", \"**\" and the rest are not read\nout, headings become their own spoken lines, links keep their text, and fenced\ncode blocks are announced instead of being spelled out.\n\nInput\n  <file>                 File to read. Use \"-\" to read stdin.\n  --text <string>        Speak this string instead of reading a file.\n  -f, --format <fmt>     auto (default), markdown, or text.\n\nOutput\n  -o, --out <file>       Audio file to write. Default: the input path with a\n                         .wav extension. Use \"-\" to write the WAV to stdout.\n  -p, --print            Print the speakable text and exit — no model, no audio.\n                         Useful for checking the Markdown conversion.\n  -q, --quiet            No progress output.\n\nVoice\n  -v, --voice <id>       Voice id. Default af_heart. See --list-voices.\n  -s, --speed <n>        Speaking rate, 0.5-2. Default 1.\n      --list-voices      Print the available voices and exit.\n\nMarkdown handling\n  --headings <mode>      text (default) | announce | skip\n  --code <mode>          announce (default) | read | skip\n  --links <mode>         text (default) | text-and-url\n  --tables <mode>        rows (default) | skip\n  --front-matter         Read the YAML front matter instead of skipping it.\n\nModel\n  --model <id>           Hugging Face model id.\n                         Default onnx-community/Kokoro-82M-v1.0-ONNX.\n  --dtype <type>         fp32 | fp16 | q8 (default) | q4 | q4f16\n  --device <device>      cpu (default) | wasm | webgpu\n  --chunk <chars>        Target characters per synthesis chunk. Default 400.\n  --gap <ms>             Silence between chunks. Default 120.\n\nOther\n  -h, --help             Show this help.\n  -V, --version          Print the package version.\n\nThe first run downloads the Kokoro weights (about 90 MB at the default q8) into\nthe Hugging Face cache; later runs are offline. Speech is synthesized locally —\nno text leaves the machine.`;\n\n/** Reads the next value for a flag, recording an error when it is missing. */\nfunction takeValue(\n  argv: string[],\n  index: number,\n  flag: string,\n  errors: string[]\n): { value?: string; next: number } {\n  const value = argv[index + 1];\n  if (value === undefined || value.startsWith(\"-\")) {\n    errors.push(`${flag} needs a value`);\n    return { next: index + 1 };\n  }\n  return { value, next: index + 1 };\n}\n\n/** Parses a numeric flag, recording an error when it is not a finite number. */\nfunction toNumber(value: string | undefined, flag: string, errors: string[]): number | undefined {\n  if (value === undefined) return undefined;\n  const parsed = Number(value);\n  if (!Number.isFinite(parsed)) {\n    errors.push(`${flag} expects a number, got \"${value}\"`);\n    return undefined;\n  }\n  return parsed;\n}\n\n/** Parses an enum flag, recording an error when the value is not in `allowed`. */\nfunction toChoice<T extends string>(\n  value: string | undefined,\n  allowed: readonly T[],\n  flag: string,\n  errors: string[]\n): T | undefined {\n  if (value === undefined) return undefined;\n  if (!allowed.includes(value as T)) {\n    errors.push(`${flag} expects one of ${allowed.join(\", \")} — got \"${value}\"`);\n    return undefined;\n  }\n  return value as T;\n}\n\n/**\n * Parses the command line. Pure: it reads nothing and writes nothing, so the\n * tests can assert on the whole option set.\n */\nexport function parseArgs(argv: string[]): CliOptions {\n  const errors: string[] = [];\n  const markdown: MarkdownToSpeechOptions = {};\n  const options: CliOptions = {\n    command: \"speak\",\n    format: \"auto\",\n    voice: \"af_heart\",\n    speed: 1,\n    markdown,\n    quiet: false,\n    errors,\n  };\n\n  let positionalsOnly = false;\n\n  for (let i = 0; i < argv.length; i += 1) {\n    const arg = argv[i];\n\n    if (positionalsOnly || !arg.startsWith(\"-\") || arg === \"-\") {\n      if (options.input !== undefined) {\n        errors.push(`unexpected extra input \"${arg}\" — pass one file at a time`);\n      } else {\n        options.input = arg;\n      }\n      continue;\n    }\n\n    if (arg === \"--\") {\n      positionalsOnly = true;\n      continue;\n    }\n\n    // `--voice=af_bella` is the same as `--voice af_bella`.\n    const equals = arg.indexOf(\"=\");\n    if (equals > 1) {\n      argv.splice(i, 1, arg.slice(0, equals), arg.slice(equals + 1));\n      i -= 1;\n      continue;\n    }\n\n    switch (arg) {\n      case \"-h\":\n      case \"--help\":\n        options.command = \"help\";\n        return options;\n\n      case \"-V\":\n      case \"--version\":\n        options.command = \"version\";\n        return options;\n\n      case \"--list-voices\":\n        options.command = \"list-voices\";\n        return options;\n\n      case \"-p\":\n      case \"--print\":\n      case \"--dry-run\":\n        options.command = \"print\";\n        break;\n\n      case \"-q\":\n      case \"--quiet\":\n        options.quiet = true;\n        break;\n\n      case \"--front-matter\":\n        markdown.frontMatter = true;\n        break;\n\n      case \"-o\":\n      case \"--out\":\n      case \"--output\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        options.output = value;\n        i = next;\n        break;\n      }\n\n      case \"--text\": {\n        // Unlike every other value, this one may legitimately start with \"-\".\n        const value = argv[i + 1];\n        if (value === undefined) errors.push(\"--text needs a value\");\n        else options.text = value;\n        i += 1;\n        break;\n      }\n\n      case \"-f\":\n      case \"--format\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        if (value !== undefined) {\n          const format = FORMATS[value.toLowerCase()];\n          if (!format) {\n            errors.push(`--format expects auto, markdown or text — got \"${value}\"`);\n          } else {\n            options.format = format;\n          }\n        }\n        break;\n      }\n\n      case \"-v\":\n      case \"--voice\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        if (value !== undefined) {\n          if (!KOKORO_VOICES.includes(value as (typeof KOKORO_VOICES)[number])) {\n            errors.push(`unknown voice \"${value}\" — run --list-voices to see them all`);\n          } else {\n            options.voice = value;\n          }\n        }\n        break;\n      }\n\n      case \"-s\":\n      case \"--speed\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        const speed = toNumber(value, arg, errors);\n        if (speed !== undefined) {\n          if (speed < 0.5 || speed > 2) errors.push(\"--speed must be between 0.5 and 2\");\n          else options.speed = speed;\n        }\n        break;\n      }\n\n      case \"--model\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        options.model = value;\n        i = next;\n        break;\n      }\n\n      case \"--dtype\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        options.dtype = toChoice(value, DTYPES, arg, errors) ?? options.dtype;\n        break;\n      }\n\n      case \"--device\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        options.device = toChoice(value, DEVICES, arg, errors) ?? options.device;\n        break;\n      }\n\n      case \"--chunk\":\n      case \"--chunk-length\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        const chunk = toNumber(value, arg, errors);\n        if (chunk !== undefined) {\n          if (chunk < 40) errors.push(\"--chunk must be at least 40 characters\");\n          else options.maxChunkLength = chunk;\n        }\n        break;\n      }\n\n      case \"--gap\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        const gap = toNumber(value, arg, errors);\n        if (gap !== undefined) {\n          if (gap < 0) errors.push(\"--gap cannot be negative\");\n          else options.gapMs = gap;\n        }\n        break;\n      }\n\n      case \"--headings\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        markdown.headings =\n          toChoice(value, [\"text\", \"announce\", \"skip\"] as const, arg, errors) ??\n          markdown.headings;\n        break;\n      }\n\n      case \"--code\":\n      case \"--code-blocks\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        markdown.codeBlocks =\n          toChoice(value, [\"announce\", \"read\", \"skip\"] as const, arg, errors) ??\n          markdown.codeBlocks;\n        break;\n      }\n\n      case \"--links\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        markdown.links =\n          toChoice(value, [\"text\", \"text-and-url\"] as const, arg, errors) ?? markdown.links;\n        break;\n      }\n\n      case \"--tables\": {\n        const { value, next } = takeValue(argv, i, arg, errors);\n        i = next;\n        markdown.tables =\n          toChoice(value, [\"rows\", \"skip\"] as const, arg, errors) ?? markdown.tables;\n        break;\n      }\n\n      default:\n        errors.push(`unknown option \"${arg}\" — run --help to see the options`);\n    }\n  }\n\n  if (options.text !== undefined && options.input !== undefined) {\n    errors.push(\"pass either a file or --text, not both\");\n  }\n\n  return options;\n}\n\n/** The `--list-voices` table, derived from the ids so it cannot drift. */\nexport function formatVoiceList(): string {\n  const rows = KOKORO_VOICES.map((id) => {\n    const voice = describeKokoroVoice(id);\n    return `  ${id.padEnd(14)}${voice.name.padEnd(12)}${voice.gender.padEnd(8)}${voice.accent}`;\n  });\n  return [`${KOKORO_VOICES.length} Kokoro voices:`, ...rows].join(\"\\n\");\n}\n\n/** Reads the package version, so `--version` matches what npx installed. */\nasync function readVersion(): Promise<string> {\n  try {\n    const { readFile } = await import(\"node:fs/promises\");\n    // From `dist/cli.js` the manifest is one level up; from a source checkout it\n    // is two. Try both before giving up.\n    for (const relative of [\"../package.json\", \"../../package.json\"]) {\n      try {\n        const url = new URL(relative, import.meta.url);\n        const manifest = JSON.parse(await readFile(url, \"utf8\"));\n        if (manifest?.name === \"use-voice-control\" && manifest.version) return manifest.version;\n      } catch {\n        // Try the next candidate.\n      }\n    }\n  } catch {\n    // No filesystem (bundled for a runtime without fs) — fall through.\n  }\n  return \"unknown\";\n}\n\n/**\n * Runs the command line and resolves to the process exit code.\n *\n * @param argv Arguments after the executable and script, i.e. `process.argv.slice(2)`.\n * @param io Injection points for output, stdin and synthesis.\n */\nexport async function runCli(argv: string[], io: CliIO = {}): Promise<number> {\n  const log = io.log ?? ((line: string) => process.stdout.write(`${line}\\n`));\n  const warn = io.error ?? ((line: string) => process.stderr.write(`${line}\\n`));\n\n  const options = parseArgs([...argv]);\n\n  if (options.command === \"help\") {\n    log(USAGE);\n    return 0;\n  }\n\n  if (options.command === \"version\") {\n    log(await readVersion());\n    return 0;\n  }\n\n  if (options.command === \"list-voices\") {\n    log(formatVoiceList());\n    return 0;\n  }\n\n  if (options.errors.length > 0) {\n    options.errors.forEach((message) => warn(`use-voice-control: ${message}`));\n    warn(\"Run `npx use-voice-control --help` for usage.\");\n    return 1;\n  }\n\n  // With nothing named on the command line, a pipe is the intended input.\n  const stdinIsTTY = io.stdinIsTTY ?? Boolean(process.stdin.isTTY);\n  const input =\n    options.input ?? (options.text === undefined && !stdinIsTTY ? \"-\" : undefined);\n\n  if (input === undefined && options.text === undefined) {\n    warn(\"use-voice-control: no input — pass a file, pipe text in, or use --text\");\n    warn(\"Run `npx use-voice-control --help` for usage.\");\n    return 1;\n  }\n\n  const shared = {\n    file: input,\n    text: options.text,\n    format: options.format,\n    markdown: options.markdown,\n    readStdin: io.readStdin,\n  };\n\n  try {\n    if (options.command === \"print\") {\n      const document = await loadDocument(shared);\n      log(document.text);\n      return 0;\n    }\n\n    const started = Date.now();\n    const result = await renderDocument({\n      ...shared,\n      output: options.output,\n      voice: options.voice,\n      speed: options.speed,\n      model: options.model,\n      dtype: options.dtype,\n      device: options.device,\n      maxChunkLength: options.maxChunkLength,\n      gapMs: options.gapMs,\n      synthesize: io.synthesize,\n      onModelProgress: options.quiet\n        ? undefined\n        : (progress: any) => {\n            if (progress?.status === \"progress\" && progress.file && progress.total) {\n              const done = Math.round(((progress.loaded ?? 0) / progress.total) * 100);\n              warn(`downloading ${progress.file}: ${done}%`);\n            }\n          },\n      onChunk: options.quiet\n        ? undefined\n        : ({ index, total }) => warn(`speaking chunk ${index + 1}/${total}`),\n    });\n\n    if (!options.quiet) {\n      const seconds = result.durationSeconds.toFixed(1);\n      const elapsed = ((Date.now() - started) / 1000).toFixed(1);\n      const target = result.output === \"-\" ? \"stdout\" : result.output;\n      warn(`wrote ${target} — ${seconds}s of audio from ${result.source} in ${elapsed}s`);\n    }\n    return 0;\n  } catch (error) {\n    warn(`use-voice-control: ${error instanceof Error ? error.message : String(error)}`);\n    return 1;\n  }\n}\n"],"names":["MARKDOWN_EXTENSIONS","TEXT_EXTENSIONS","extensionOf","filename","base","dot","detectFormat","content","requested","extension","looksLikeMarkdown","toSpeechText","format","options","markdownToSpeech","loadDocument","file","text","markdown","resolved","stdin","readStdinToString","readFile","chunks","chunk","defaultOutputPath","input","renderDocument","document","synthesize","synthesizeSamples","samples","sampleRate","audio","encodeWav","output","writeFile","wavDurationSeconds","FORMATS","DTYPES","DEVICES","USAGE","takeValue","argv","index","flag","errors","value","toNumber","parsed","toChoice","allowed","parseArgs","positionalsOnly","arg","equals","next","KOKORO_VOICES","speed","gap","formatVoiceList","rows","id","voice","describeKokoroVoice","readVersion","relative","url","manifest","runCli","io","log","line","warn","message","stdinIsTTY","shared","started","result","progress","done","total","seconds","elapsed","target","error"],"mappings":";;AAmBO,MAAMA,IAAsB,CAAC,OAAO,aAAa,UAAU,QAAQ,MAAM,GAGnEC,IAAkB,CAAC,QAAQ,SAAS,EAAE;AAG5C,SAASC,EAAYC,GAA0B;AACpD,QAAMC,IAAOD,EAAS,MAAM,OAAO,EAAE,SAAS,IACxCE,IAAMD,EAAK,YAAY,GAAG;AAChC,SAAOC,IAAM,IAAID,EAAK,MAAMC,CAAG,EAAE,gBAAgB;AACnD;AASO,SAASC,EACdC,GACAJ,GACAK,IAA6B,QACb;AAChB,MAAIA,MAAc,OAAQ,QAAOA;AAEjC,MAAIL,GAAU;AACZ,UAAMM,IAAYP,EAAYC,CAAQ;AACtC,QAAIH,EAAoB,SAASS,CAAS,EAAG,QAAO;AACpD,QAAIR,EAAgB,SAASQ,CAAS,EAAG,QAAO;AAAA,EAClD;AAEA,SAAOC,EAAkBH,CAAO,IAAI,aAAa;AACnD;AAGO,SAASI,EACdJ,GACAK,GACAC,IAAmC,CAAA,GAC3B;AACR,SAAID,MAAW,SAGNL,EAAQ,QAAQ,UAAU;AAAA,CAAI,EAAE,KAAA,IAElCO,EAAiBP,GAASM,CAAO;AAC1C;AA0BA,eAAsBE,EACpBF,GACyB;AACzB,QAAM,EAAE,MAAAG,GAAM,MAAAC,GAAM,QAAAL,IAAS,QAAQ,UAAAM,IAAW,CAAA,MAAOL;AAEvD,MAAII,MAAS,QAAW;AACtB,UAAME,IAAWb,EAAaW,GAAM,QAAWL,CAAM;AACrD,WAAO,EAAE,MAAMD,EAAaM,GAAME,GAAUD,CAAQ,GAAG,QAAQC,GAAU,QAAQ,SAAA;AAAA,EACnF;AAEA,MAAI,CAACH,EAAM,OAAM,IAAI,MAAM,sDAAsD;AAEjF,MAAIA,MAAS,KAAK;AAChB,UAAMI,IAAQP,EAAQ,YAClB,MAAMA,EAAQ,UAAA,IACd,MAAMQ,EAAA,GACJF,IAAWb,EAAac,GAAO,QAAWR,CAAM;AACtD,WAAO,EAAE,MAAMD,EAAaS,GAAOD,GAAUD,CAAQ,GAAG,QAAQC,GAAU,QAAQ,QAAA;AAAA,EACpF;AAEA,QAAM,EAAE,UAAAG,EAAA,IAAa,MAAM,OAAO,kBAAkB,GAC9Cf,IAAU,MAAMe,EAASN,GAAM,MAAM,GACrCG,IAAWb,EAAaC,GAASS,GAAMJ,CAAM;AACnD,SAAO,EAAE,MAAMD,EAAaJ,GAASY,GAAUD,CAAQ,GAAG,QAAQC,GAAU,QAAQH,EAAA;AACtF;AAGA,eAAsBK,IAAqC;AACzD,QAAME,IAAmB,CAAA;AACzB,mBAAiBC,KAAS,QAAQ;AAChC,IAAAD,EAAO,KAAK,OAAO,SAASC,CAAK,IAAIA,IAAQ,OAAO,KAAKA,CAAK,CAAC;AAEjE,SAAO,OAAO,OAAOD,CAAM,EAAE,SAAS,MAAM;AAC9C;ACzFO,SAASE,EAAkBC,GAAuB;AACvD,QAAMjB,IAAYP,EAAYwB,CAAK;AAEnC,SAAO,GADMjB,IAAYiB,EAAM,MAAM,GAAG,CAACjB,EAAU,MAAM,IAAIiB,CAC/C;AAChB;AAQA,eAAsBC,EAAed,GAA+C;AAClF,QAAMe,IAAW,MAAMb,EAAaF,CAAO;AAE3C,MAAI,CAACe,EAAS,KAAK;AACjB,UAAM,IAAI,MAAM,qBAAqBA,EAAS,MAAM,uBAAuB;AAG7E,QAAMC,IAAahB,EAAQ,cAAciB,GACnC,EAAE,SAAAC,GAAS,YAAAC,EAAA,IAAe,MAAMH,EAAWD,EAAS,MAAMf,CAAO,GACjEoB,IAAQC,EAAUH,GAASC,CAAU,GAErCG,IACJtB,EAAQ,WACPA,EAAQ,QAAQA,EAAQ,SAAS,MAAMY,EAAkBZ,EAAQ,IAAI,IAAI;AAE5E,MAAIsB,MAAW;AACb,YAAQ,OAAO,MAAM,OAAO,KAAKF,CAAK,CAAC;AAAA,OAClC;AACL,UAAM,EAAE,WAAAG,EAAA,IAAc,MAAM,OAAO,kBAAkB;AACrD,UAAMA,EAAUD,GAAQ,OAAO,KAAKF,CAAK,CAAC;AAAA,EAC5C;AAEA,SAAO;AAAA,IACL,QAAAE;AAAA,IACA,MAAMP,EAAS;AAAA,IACf,QAAQA,EAAS;AAAA,IACjB,QAAQA,EAAS;AAAA,IACjB,iBAAiBS,EAAmBN,GAASC,CAAU;AAAA,IACvD,OAAOC,EAAM;AAAA,EAAA;AAEjB;ACvBA,MAAMK,IAA2C;AAAA,EAC/C,MAAM;AAAA,EACN,UAAU;AAAA,EACV,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,KAAK;AAAA,EACL,OAAO;AACT,GAEMC,IAAwB,CAAC,QAAQ,QAAQ,MAAM,MAAM,OAAO,GAC5DC,IAA0B,CAAC,QAAQ,UAAU,KAAK,GAE3CC,IAAQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoDrB,SAASC,EACPC,GACAC,GACAC,GACAC,GACkC;AAClC,QAAMC,IAAQJ,EAAKC,IAAQ,CAAC;AAC5B,SAAIG,MAAU,UAAaA,EAAM,WAAW,GAAG,KAC7CD,EAAO,KAAK,GAAGD,CAAI,gBAAgB,GAC5B,EAAE,MAAMD,IAAQ,EAAA,KAElB,EAAE,OAAAG,GAAO,MAAMH,IAAQ,EAAA;AAChC;AAGA,SAASI,EAASD,GAA2BF,GAAcC,GAAsC;AAC/F,MAAIC,MAAU,OAAW;AACzB,QAAME,IAAS,OAAOF,CAAK;AAC3B,MAAI,CAAC,OAAO,SAASE,CAAM,GAAG;AAC5B,IAAAH,EAAO,KAAK,GAAGD,CAAI,2BAA2BE,CAAK,GAAG;AACtD;AAAA,EACF;AACA,SAAOE;AACT;AAGA,SAASC,EACPH,GACAI,GACAN,GACAC,GACe;AACf,MAAIC,MAAU,QACd;AAAA,QAAI,CAACI,EAAQ,SAASJ,CAAU,GAAG;AACjC,MAAAD,EAAO,KAAK,GAAGD,CAAI,mBAAmBM,EAAQ,KAAK,IAAI,CAAC,WAAWJ,CAAK,GAAG;AAC3E;AAAA,IACF;AACA,WAAOA;AAAA;AACT;AAMO,SAASK,EAAUT,GAA4B;AACpD,QAAMG,IAAmB,CAAA,GACnB5B,IAAoC,CAAA,GACpCL,IAAsB;AAAA,IAC1B,SAAS;AAAA,IACT,QAAQ;AAAA,IACR,OAAO;AAAA,IACP,OAAO;AAAA,IACP,UAAAK;AAAA,IACA,OAAO;AAAA,IACP,QAAA4B;AAAA,EAAA;AAGF,MAAIO,IAAkB;AAEtB,WAAS,IAAI,GAAG,IAAIV,EAAK,QAAQ,KAAK,GAAG;AACvC,UAAMW,IAAMX,EAAK,CAAC;AAElB,QAAIU,KAAmB,CAACC,EAAI,WAAW,GAAG,KAAKA,MAAQ,KAAK;AAC1D,MAAIzC,EAAQ,UAAU,SACpBiC,EAAO,KAAK,2BAA2BQ,CAAG,6BAA6B,IAEvEzC,EAAQ,QAAQyC;AAElB;AAAA,IACF;AAEA,QAAIA,MAAQ,MAAM;AAChB,MAAAD,IAAkB;AAClB;AAAA,IACF;AAGA,UAAME,IAASD,EAAI,QAAQ,GAAG;AAC9B,QAAIC,IAAS,GAAG;AACd,MAAAZ,EAAK,OAAO,GAAG,GAAGW,EAAI,MAAM,GAAGC,CAAM,GAAGD,EAAI,MAAMC,IAAS,CAAC,CAAC,GAC7D,KAAK;AACL;AAAA,IACF;AAEA,YAAQD,GAAA;AAAA,MACN,KAAK;AAAA,MACL,KAAK;AACH,eAAAzC,EAAQ,UAAU,QACXA;AAAA,MAET,KAAK;AAAA,MACL,KAAK;AACH,eAAAA,EAAQ,UAAU,WACXA;AAAA,MAET,KAAK;AACH,eAAAA,EAAQ,UAAU,eACXA;AAAA,MAET,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AACH,QAAAA,EAAQ,UAAU;AAClB;AAAA,MAEF,KAAK;AAAA,MACL,KAAK;AACH,QAAAA,EAAQ,QAAQ;AAChB;AAAA,MAEF,KAAK;AACH,QAAAK,EAAS,cAAc;AACvB;AAAA,MAEF,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK,YAAY;AACf,cAAM,EAAE,OAAA6B,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,QAAAjC,EAAQ,SAASkC,GACjB,IAAIS;AACJ;AAAA,MACF;AAAA,MAEA,KAAK,UAAU;AAEb,cAAMT,IAAQJ,EAAK,IAAI,CAAC;AACxB,QAAII,MAAU,SAAWD,EAAO,KAAK,sBAAsB,MAC9C,OAAOC,GACpB,KAAK;AACL;AAAA,MACF;AAAA,MAEA,KAAK;AAAA,MACL,KAAK,YAAY;AACf,cAAM,EAAE,OAAAA,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AAEtD,YADA,IAAIU,GACAT,MAAU,QAAW;AACvB,gBAAMnC,IAAS0B,EAAQS,EAAM,YAAA,CAAa;AAC1C,UAAKnC,IAGHC,EAAQ,SAASD,IAFjBkC,EAAO,KAAK,kDAAkDC,CAAK,GAAG;AAAA,QAI1E;AACA;AAAA,MACF;AAAA,MAEA,KAAK;AAAA,MACL,KAAK,WAAW;AACd,cAAM,EAAE,OAAAA,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACAT,MAAU,WACPU,EAAc,SAASV,CAAuC,IAGjElC,EAAQ,QAAQkC,IAFhBD,EAAO,KAAK,kBAAkBC,CAAK,uCAAuC;AAK9E;AAAA,MACF;AAAA,MAEA,KAAK;AAAA,MACL,KAAK,WAAW;AACd,cAAM,EAAE,OAAAA,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU;AACJ,cAAME,IAAQV,EAASD,GAAOO,GAAKR,CAAM;AACzC,QAAIY,MAAU,WACRA,IAAQ,OAAOA,IAAQ,IAAGZ,EAAO,KAAK,mCAAmC,MAChE,QAAQY;AAEvB;AAAA,MACF;AAAA,MAEA,KAAK,WAAW;AACd,cAAM,EAAE,OAAAX,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,QAAAjC,EAAQ,QAAQkC,GAChB,IAAIS;AACJ;AAAA,MACF;AAAA,MAEA,KAAK,WAAW;AACd,cAAM,EAAE,OAAAT,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJ3C,EAAQ,QAAQqC,EAASH,GAAOR,GAAQe,GAAKR,CAAM,KAAKjC,EAAQ;AAChE;AAAA,MACF;AAAA,MAEA,KAAK,YAAY;AACf,cAAM,EAAE,OAAAkC,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJ3C,EAAQ,SAASqC,EAASH,GAAOP,GAASc,GAAKR,CAAM,KAAKjC,EAAQ;AAClE;AAAA,MACF;AAAA,MAEA,KAAK;AAAA,MACL,KAAK,kBAAkB;AACrB,cAAM,EAAE,OAAAkC,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU;AACJ,cAAMhC,IAAQwB,EAASD,GAAOO,GAAKR,CAAM;AACzC,QAAItB,MAAU,WACRA,IAAQ,KAAIsB,EAAO,KAAK,wCAAwC,MACvD,iBAAiBtB;AAEhC;AAAA,MACF;AAAA,MAEA,KAAK,SAAS;AACZ,cAAM,EAAE,OAAAuB,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU;AACJ,cAAMG,IAAMX,EAASD,GAAOO,GAAKR,CAAM;AACvC,QAAIa,MAAQ,WACNA,IAAM,IAAGb,EAAO,KAAK,0BAA0B,MACtC,QAAQa;AAEvB;AAAA,MACF;AAAA,MAEA,KAAK,cAAc;AACjB,cAAM,EAAE,OAAAZ,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJtC,EAAS,WACPgC,EAASH,GAAO,CAAC,QAAQ,YAAY,MAAM,GAAYO,GAAKR,CAAM,KAClE5B,EAAS;AACX;AAAA,MACF;AAAA,MAEA,KAAK;AAAA,MACL,KAAK,iBAAiB;AACpB,cAAM,EAAE,OAAA6B,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJtC,EAAS,aACPgC,EAASH,GAAO,CAAC,YAAY,QAAQ,MAAM,GAAYO,GAAKR,CAAM,KAClE5B,EAAS;AACX;AAAA,MACF;AAAA,MAEA,KAAK,WAAW;AACd,cAAM,EAAE,OAAA6B,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJtC,EAAS,QACPgC,EAASH,GAAO,CAAC,QAAQ,cAAc,GAAYO,GAAKR,CAAM,KAAK5B,EAAS;AAC9E;AAAA,MACF;AAAA,MAEA,KAAK,YAAY;AACf,cAAM,EAAE,OAAA6B,GAAO,MAAAS,MAASd,EAAUC,GAAM,GAAGW,GAAKR,CAAM;AACtD,YAAIU,GACJtC,EAAS,SACPgC,EAASH,GAAO,CAAC,QAAQ,MAAM,GAAYO,GAAKR,CAAM,KAAK5B,EAAS;AACtE;AAAA,MACF;AAAA,MAEA;AACE,QAAA4B,EAAO,KAAK,mBAAmBQ,CAAG,mCAAmC;AAAA,IAAA;AAAA,EAE3E;AAEA,SAAIzC,EAAQ,SAAS,UAAaA,EAAQ,UAAU,UAClDiC,EAAO,KAAK,wCAAwC,GAG/CjC;AACT;AAGO,SAAS+C,IAA0B;AACxC,QAAMC,IAAOJ,EAAc,IAAI,CAACK,MAAO;AACrC,UAAMC,IAAQC,EAAoBF,CAAE;AACpC,WAAO,KAAKA,EAAG,OAAO,EAAE,CAAC,GAAGC,EAAM,KAAK,OAAO,EAAE,CAAC,GAAGA,EAAM,OAAO,OAAO,CAAC,CAAC,GAAGA,EAAM,MAAM;AAAA,EAC3F,CAAC;AACD,SAAO,CAAC,GAAGN,EAAc,MAAM,mBAAmB,GAAGI,CAAI,EAAE,KAAK;AAAA,CAAI;AACtE;AAGA,eAAeI,IAA+B;AAC5C,MAAI;AACF,UAAM,EAAE,UAAA3C,EAAA,IAAa,MAAM,OAAO,kBAAkB;AAGpD,eAAW4C,KAAY,CAAC,mBAAmB,oBAAoB;AAC7D,UAAI;AACF,cAAMC,IAAM,IAAI,IAAID,GAAU,YAAY,GAAG,GACvCE,IAAW,KAAK,MAAM,MAAM9C,EAAS6C,GAAK,MAAM,CAAC;AACvD,aAAIC,KAAA,gBAAAA,EAAU,UAAS,uBAAuBA,EAAS,gBAAgBA,EAAS;AAAA,MAClF,QAAQ;AAAA,MAER;AAAA,EAEJ,QAAQ;AAAA,EAER;AACA,SAAO;AACT;AAQA,eAAsBC,EAAO1B,GAAgB2B,IAAY,IAAqB;AAC5E,QAAMC,IAAMD,EAAG,QAAQ,CAACE,MAAiB,QAAQ,OAAO,MAAM,GAAGA,CAAI;AAAA,CAAI,IACnEC,IAAOH,EAAG,UAAU,CAACE,MAAiB,QAAQ,OAAO,MAAM,GAAGA,CAAI;AAAA,CAAI,IAEtE3D,IAAUuC,EAAU,CAAC,GAAGT,CAAI,CAAC;AAEnC,MAAI9B,EAAQ,YAAY;AACtB,WAAA0D,EAAI9B,CAAK,GACF;AAGT,MAAI5B,EAAQ,YAAY;AACtB,WAAA0D,EAAI,MAAMN,GAAa,GAChB;AAGT,MAAIpD,EAAQ,YAAY;AACtB,WAAA0D,EAAIX,GAAiB,GACd;AAGT,MAAI/C,EAAQ,OAAO,SAAS;AAC1B,WAAAA,EAAQ,OAAO,QAAQ,CAAC6D,MAAYD,EAAK,sBAAsBC,CAAO,EAAE,CAAC,GACzED,EAAK,+CAA+C,GAC7C;AAIT,QAAME,IAAaL,EAAG,cAAc,EAAQ,QAAQ,MAAM,OACpD5C,IACJb,EAAQ,UAAUA,EAAQ,SAAS,UAAa,CAAC8D,IAAa,MAAM;AAEtE,MAAIjD,MAAU,UAAab,EAAQ,SAAS;AAC1C,WAAA4D,EAAK,wEAAwE,GAC7EA,EAAK,+CAA+C,GAC7C;AAGT,QAAMG,IAAS;AAAA,IACb,MAAMlD;AAAA,IACN,MAAMb,EAAQ;AAAA,IACd,QAAQA,EAAQ;AAAA,IAChB,UAAUA,EAAQ;AAAA,IAClB,WAAWyD,EAAG;AAAA,EAAA;AAGhB,MAAI;AACF,QAAIzD,EAAQ,YAAY,SAAS;AAC/B,YAAMe,IAAW,MAAMb,EAAa6D,CAAM;AAC1C,aAAAL,EAAI3C,EAAS,IAAI,GACV;AAAA,IACT;AAEA,UAAMiD,IAAU,KAAK,IAAA,GACfC,IAAS,MAAMnD,EAAe;AAAA,MAClC,GAAGiD;AAAA,MACH,QAAQ/D,EAAQ;AAAA,MAChB,OAAOA,EAAQ;AAAA,MACf,OAAOA,EAAQ;AAAA,MACf,OAAOA,EAAQ;AAAA,MACf,OAAOA,EAAQ;AAAA,MACf,QAAQA,EAAQ;AAAA,MAChB,gBAAgBA,EAAQ;AAAA,MACxB,OAAOA,EAAQ;AAAA,MACf,YAAYyD,EAAG;AAAA,MACf,iBAAiBzD,EAAQ,QACrB,SACA,CAACkE,MAAkB;AACjB,aAAIA,KAAA,gBAAAA,EAAU,YAAW,cAAcA,EAAS,QAAQA,EAAS,OAAO;AACtE,gBAAMC,IAAO,KAAK,OAAQD,EAAS,UAAU,KAAKA,EAAS,QAAS,GAAG;AACvE,UAAAN,EAAK,eAAeM,EAAS,IAAI,KAAKC,CAAI,GAAG;AAAA,QAC/C;AAAA,MACF;AAAA,MACJ,SAASnE,EAAQ,QACb,SACA,CAAC,EAAE,OAAA+B,GAAO,OAAAqC,EAAA,MAAYR,EAAK,kBAAkB7B,IAAQ,CAAC,IAAIqC,CAAK,EAAE;AAAA,IAAA,CACtE;AAED,QAAI,CAACpE,EAAQ,OAAO;AAClB,YAAMqE,IAAUJ,EAAO,gBAAgB,QAAQ,CAAC,GAC1CK,MAAY,KAAK,IAAA,IAAQN,KAAW,KAAM,QAAQ,CAAC,GACnDO,IAASN,EAAO,WAAW,MAAM,WAAWA,EAAO;AACzD,MAAAL,EAAK,SAASW,CAAM,MAAMF,CAAO,mBAAmBJ,EAAO,MAAM,OAAOK,CAAO,GAAG;AAAA,IACpF;AACA,WAAO;AAAA,EACT,SAASE,GAAO;AACd,WAAAZ,EAAK,sBAAsBY,aAAiB,QAAQA,EAAM,UAAU,OAAOA,CAAK,CAAC,EAAE,GAC5E;AAAA,EACT;AACF;"}

package/dist/client/index.d.ts

@@ -5,4 +5,5 @@
  */
 export { ReadAloudController, type ReadAloudChunk, type ReadAloudOptions, type ReadAloudState, type SynthesizeFn, } from './read-aloud';
 export { LiveTranscriber, isTranscriptionSupported, type LiveTranscriberOptions, type TranscriberEngine, } from './live-transcriber';
+export { looksLikeMarkdown, markdownToSpeech, markdownToSpeechSegments, stripInlineMarkdown, type MarkdownToSpeechOptions, type SpeechSegment, type SpeechSegmentType, } from '../utils/markdown-to-speech';
 export type { TTSProvider, KokoroVoice, DeepgramSpeaker } from '../types/types';

package/dist/client/read-aloud.d.ts

@@ -1,4 +1,5 @@
 import { TTSProvider } from '../types/types';
+import { MarkdownToSpeechOptions } from '../utils/markdown-to-speech';
 export type ReadAloudState = "idle" | "loading" | "speaking" | "paused";
 export interface ReadAloudChunk {
     /** The text of the chunk currently being spoken. */
@@ -22,6 +23,14 @@ export interface ReadAloudOptions {
      * seams between chunks more audible. Default 240.
      */
     maxChunkLength?: number;
+    /**
+     * How to read the text handed to `speak()`. `auto` (default) converts text
+     * that looks like Markdown so "##" and "**" are not read out; `markdown`
+     * always converts; `text` never does.
+     */
+    format?: "auto" | "markdown" | "text";
+    /** Markdown conversion options, when the text is read as Markdown. */
+    markdown?: MarkdownToSpeechOptions;
     /** Override synthesis entirely (tests, a bring-your-own-TTS host app). */
     synthesize?: SynthesizeFn;
     /** Called as each chunk starts playing. */
@@ -55,6 +64,11 @@ export declare class ReadAloudController {
     resume(): void;
     /** Stop playback and drop any queued chunks. Safe to call when already idle. */
     stop(): void;
+    /**
+     * Converts Markdown to spoken words before chunking, so a document read out
+     * of an editor does not have its syntax read back to the listener.
+     */
+    private toSpeakableText;
     private setState;
     private cleanup;
     private synthesize;

package/dist/client.js

@@ -1,66 +1,29 @@
-var U = Object.defineProperty;
-var E = (h, e, i) => e in h ? U(h, e, { enumerable: !0, configurable: !0, writable: !0, value: i }) : h[e] = i;
-var l = (h, e, i) => E(h, typeof e != "symbol" ? e + "" : e, i);
-function C(h, e = 500) {
-  const i = h.split(/\n\s*\n/), t = [];
-  for (let s of i) {
-    if (s.length <= e) {
-      t.push(s.trim());
-      continue;
-    }
-    const n = new RegExp(`(?<=[.?!])(?=\\s+["“”'a-z])`, "gi"), a = s.split(n);
-    let o = "";
-    for (let r of a) {
-      if (r = r.trim(), r.length > e) {
-        const p = k(r, e);
-        for (let c of p)
-          (o + " " + c).length > e ? (o && t.push(o.trim()), o = c) : o += (o ? " " : "") + c;
-        continue;
-      }
-      (o + " " + r).length > e ? (o && t.push(o.trim()), o = r) : o += (o ? " " : "") + r;
-    }
-    o && t.push(o.trim());
-  }
-  return t;
-}
-function k(h, e) {
-  const i = [];
-  let t = "";
-  const s = h.split(/,\s*/);
-  for (let n of s)
-    if ((t + ", " + n).length > e)
-      if (t && i.push(t.trim()), n.length > e) {
-        const a = n.split(/\s+/);
-        let o = "";
-        for (let r of a)
-          (o + " " + r).length > e ? (o && i.push(o.trim()), o = r) : o += (o ? " " : "") + r;
-        o && i.push(o.trim()), t = "";
-      } else
-        t = n;
-    else
-      t += (t ? ", " : "") + n;
-  return t && i.push(t.trim()), i;
-}
-const v = "/api/speech/tts", R = "af_heart", L = 240;
-function w(h) {
-  return h.aborted ? Promise.resolve() : new Promise((e) => {
-    h.addEventListener("abort", () => e(), { once: !0 });
+var k = Object.defineProperty;
+var U = (r, t, e) => t in r ? k(r, t, { enumerable: !0, configurable: !0, writable: !0, value: e }) : r[t] = e;
+var a = (r, t, e) => U(r, typeof t != "symbol" ? t + "" : t, e);
+import { looksLikeMarkdown as E, markdownToSpeech as L } from "./markdown.js";
+import { markdownToSpeechSegments as D, stripInlineMarkdown as N } from "./markdown.js";
+import { s as C } from "./semantic-split-CXhk-k1F.js";
+const T = "/api/speech/tts", v = "af_heart", R = 240;
+function w(r) {
+  return r.aborted ? Promise.resolve() : new Promise((t) => {
+    r.addEventListener("abort", () => t(), { once: !0 });
   });
 }
-class A {
-  constructor(e = {}) {
-    l(this, "options");
-    l(this, "state", "idle");
-    l(this, "abortController", null);
-    l(this, "audio", null);
-    l(this, "objectUrl", null);
+class O {
+  constructor(t = {}) {
+    a(this, "options");
+    a(this, "state", "idle");
+    a(this, "abortController", null);
+    a(this, "audio", null);
+    a(this, "objectUrl", null);
     /** Set once the endpoint has proved unreachable, so later chunks skip the retry. */
-    l(this, "endpointUnavailable", !1);
-    this.options = e;
+    a(this, "endpointUnavailable", !1);
+    this.options = t;
   }
   /** Replace the options (voice, callbacks, …) without discarding playback state. */
-  setOptions(e) {
-    this.options = e;
+  setOptions(t) {
+    this.options = t;
   }
   getState() {
     return this.state;
@@ -72,29 +35,29 @@ class A {
    * Speak `text`, cancelling anything already playing. Resolves when playback
    * finishes or is stopped — it never rejects; failures go to `onError`.
    */
-  async speak(e) {
-    var a, o, r, p, c, g, d, y;
+  async speak(t) {
+    var o, h, c, f, u, g, d, y;
     this.stop();
-    const i = this.options.maxChunkLength ?? L, t = C(e ?? "", i).map((u) => u.trim()).filter((u) => u.length > 0);
-    if (t.length === 0) return;
+    const e = this.options.maxChunkLength ?? R, i = C(this.toSpeakableText(t ?? ""), e).map((l) => l.trim()).filter((l) => l.length > 0);
+    if (i.length === 0) return;
     const s = new AbortController();
     this.abortController = s;
     const { signal: n } = s;
     this.setState("loading");
     try {
-      let u = this.synthesize(t[0], n);
-      for (let f = 0; f < t.length; f += 1) {
-        const m = u;
-        u = f + 1 < t.length ? this.synthesize(t[f + 1], n) : Promise.resolve(null);
-        const S = await m;
-        if (n.aborted || (this.setState("speaking"), (o = (a = this.options).onChunk) == null || o.call(a, { text: t[f], index: f, total: t.length }), n.aborted) || (await this.playChunk(S, t[f], n), n.aborted)) return;
+      let l = this.synthesize(i[0], n);
+      for (let p = 0; p < i.length; p += 1) {
+        const S = l;
+        l = p + 1 < i.length ? this.synthesize(i[p + 1], n) : Promise.resolve(null);
+        const b = await S;
+        if (n.aborted || (this.setState("speaking"), (h = (o = this.options).onChunk) == null || h.call(o, { text: i[p], index: p, total: i.length }), n.aborted) || (await this.playChunk(b, i[p], n), n.aborted)) return;
       }
-      this.cleanup(), this.setState("idle"), (p = (r = this.options).onEnd) == null || p.call(r, "finished");
-    } catch (u) {
+      this.cleanup(), this.setState("idle"), (f = (c = this.options).onEnd) == null || f.call(c, "finished");
+    } catch (l) {
       if (n.aborted) return;
-      this.cleanup(), this.setState("idle"), (g = (c = this.options).onError) == null || g.call(
-        c,
-        u instanceof Error ? u : new Error(String(u))
+      this.cleanup(), this.setState("idle"), (g = (u = this.options).onError) == null || g.call(
+        u,
+        l instanceof Error ? l : new Error(String(l))
       ), (y = (d = this.options).onEnd) == null || y.call(d, "stopped");
     } finally {
       this.abortController === s && (this.abortController = null);
@@ -109,117 +72,125 @@ class A {
   }
   /** Stop playback and drop any queued chunks. Safe to call when already idle. */
   stop() {
-    var i, t, s;
-    const e = this.isActive();
-    (i = this.abortController) == null || i.abort(), this.abortController = null, this.cleanup(), e && (this.setState("idle"), (s = (t = this.options).onEnd) == null || s.call(t, "stopped"));
+    var e, i, s;
+    const t = this.isActive();
+    (e = this.abortController) == null || e.abort(), this.abortController = null, this.cleanup(), t && (this.setState("idle"), (s = (i = this.options).onEnd) == null || s.call(i, "stopped"));
   }
-  setState(e) {
-    var i, t;
-    this.state !== e && (this.state = e, (t = (i = this.options).onStateChange) == null || t.call(i, e));
+  /**
+   * Converts Markdown to spoken words before chunking, so a document read out
+   * of an editor does not have its syntax read back to the listener.
+   */
+  toSpeakableText(t) {
+    const e = this.options.format ?? "auto";
+    return e === "text" ? t : e === "markdown" || E(t) ? L(t, this.options.markdown) : t;
+  }
+  setState(t) {
+    var e, i;
+    this.state !== t && (this.state = t, (i = (e = this.options).onStateChange) == null || i.call(e, t));
   }
   cleanup() {
     this.audio && (this.audio.pause(), this.audio.src = "", this.audio = null), this.objectUrl && (URL.revokeObjectURL(this.objectUrl), this.objectUrl = null), typeof speechSynthesis < "u" && speechSynthesis.cancel();
   }
-  synthesize(e, i) {
-    const t = this.options.synthesize ? this.options.synthesize(e, i) : this.fetchAudio(e, i);
-    return t.catch(() => {
-    }), t;
+  synthesize(t, e) {
+    const i = this.options.synthesize ? this.options.synthesize(t, e) : this.fetchAudio(t, e);
+    return i.catch(() => {
+    }), i;
   }
-  async fetchAudio(e, i) {
+  async fetchAudio(t, e) {
     if (this.endpointUnavailable) return null;
-    const t = this.options.endpoint ?? v;
+    const i = this.options.endpoint ?? T;
     try {
-      const s = await fetch(t, {
+      const s = await fetch(i, {
         method: "POST",
         headers: { "Content-Type": "application/json" },
         body: JSON.stringify({
-          text: e,
+          text: t,
           provider: this.options.provider ?? "kokoro",
-          voice: this.options.voice ?? R
+          voice: this.options.voice ?? v
         }),
-        signal: i
+        signal: e
       });
       if (!s.ok) throw new Error(`TTS failed: ${s.status}`);
       const n = s.headers.get("Content-Type") || "audio/wav";
       return new Blob([await s.arrayBuffer()], { type: n });
     } catch (s) {
-      if (i.aborted) throw s;
+      if (e.aborted) throw s;
       return this.endpointUnavailable = !0, null;
     }
   }
-  playChunk(e, i, t) {
-    return e ? this.playAudioBlob(e, t) : this.playWithSpeechSynthesis(i, t);
+  playChunk(t, e, i) {
+    return t ? this.playAudioBlob(t, i) : this.playWithSpeechSynthesis(e, i);
   }
-  async playAudioBlob(e, i) {
-    const t = URL.createObjectURL(e), s = new Audio(t);
-    this.audio = s, this.objectUrl = t;
-    const n = new Promise((a, o) => {
-      s.addEventListener("ended", () => a(), { once: !0 }), s.addEventListener(
+  async playAudioBlob(t, e) {
+    const i = URL.createObjectURL(t), s = new Audio(i);
+    this.audio = s, this.objectUrl = i;
+    const n = new Promise((o, h) => {
+      s.addEventListener("ended", () => o(), { once: !0 }), s.addEventListener(
         "error",
-        () => o(new Error("Audio playback failed")),
+        () => h(new Error("Audio playback failed")),
         { once: !0 }
       );
     });
     try {
-      await s.play(), await Promise.race([n, w(i)]);
+      await s.play(), await Promise.race([n, w(e)]);
     } finally {
-      this.audio === s && (this.audio = null), this.objectUrl === t && (this.objectUrl = null), s.pause(), URL.revokeObjectURL(t);
+      this.audio === s && (this.audio = null), this.objectUrl === i && (this.objectUrl = null), s.pause(), URL.revokeObjectURL(i);
     }
   }
-  async playWithSpeechSynthesis(e, i) {
+  async playWithSpeechSynthesis(t, e) {
     if (typeof speechSynthesis > "u" || typeof SpeechSynthesisUtterance > "u")
       throw new Error("No speech synthesis available in this browser");
-    const t = new SpeechSynthesisUtterance(e), s = new Promise((n, a) => {
-      t.onend = () => n(), t.onerror = () => i.aborted ? n() : a(new Error("Speech synthesis failed"));
+    const i = new SpeechSynthesisUtterance(t), s = new Promise((n, o) => {
+      i.onend = () => n(), i.onerror = () => e.aborted ? n() : o(new Error("Speech synthesis failed"));
     });
-    speechSynthesis.speak(t), await Promise.race([s, w(i)]);
+    speechSynthesis.speak(i), await Promise.race([s, w(e)]);
   }
 }
-function b() {
+function m() {
   return typeof window > "u" ? null : window.SpeechRecognition || window.webkitSpeechRecognition || null;
 }
-function P() {
-  var h;
-  return typeof window > "u" ? !1 : b() ? !0 : !!((h = navigator.mediaDevices) != null && h.getUserMedia);
+function x() {
+  var r;
+  return typeof window > "u" ? !1 : m() ? !0 : !!((r = navigator.mediaDevices) != null && r.getUserMedia);
 }
-class j {
-  constructor(e = {}) {
-    l(this, "options");
-    l(this, "listening", !1);
+class M {
+  constructor(t = {}) {
+    a(this, "options");
+    a(this, "listening", !1);
     /** Distinguishes a deliberate `stop()` from Chromium's idle auto-stop. */
-    l(this, "stopRequested", !1);
-    l(this, "recognition", null);
-    l(this, "moonshine", null);
-    this.options = e;
+    a(this, "stopRequested", !1);
+    a(this, "recognition", null);
+    a(this, "moonshine", null);
+    this.options = t;
   }
-  setOptions(e) {
-    this.options = e;
+  setOptions(t) {
+    this.options = t;
   }
   isListening() {
     return this.listening;
   }
   async start() {
-    var t, s;
+    var i, s;
     if (this.listening) return;
     this.stopRequested = !1;
-    const e = this.options.engine ?? "auto", i = b();
+    const t = this.options.engine ?? "auto", e = m();
     try {
-      if (e !== "moonshine" && i) {
-        this.startWebSpeech(i);
+      if (t !== "moonshine" && e) {
+        this.startWebSpeech(e);
         return;
       }
-      if (e === "webspeech")
+      if (t === "webspeech")
         throw new Error("Speech recognition is not available in this browser");
       await this.startMoonshine();
     } catch (n) {
-      this.setListening(!1), (s = (t = this.options).onError) == null || s.call(
-        t,
+      this.setListening(!1), (s = (i = this.options).onError) == null || s.call(
+        i,
         n instanceof Error ? n : new Error(String(n))
       );
     }
   }
   async stop() {
-    var e, i;
+    var t, e;
     if (this.stopRequested = !0, this.recognition) {
       try {
         this.recognition.stop();
@@ -229,7 +200,7 @@ class j {
     }
     if (this.moonshine) {
       try {
-        await ((i = (e = this.moonshine).stop) == null ? void 0 : i.call(e));
+        await ((e = (t = this.moonshine).stop) == null ? void 0 : e.call(t));
       } catch {
       }
       this.moonshine = null;
@@ -239,54 +210,54 @@ class j {
   async toggle() {
     this.listening ? await this.stop() : await this.start();
   }
-  setListening(e) {
-    var i, t;
-    this.listening !== e && (this.listening = e, (t = (i = this.options).onStateChange) == null || t.call(i, e));
+  setListening(t) {
+    var e, i;
+    this.listening !== t && (this.listening = t, (i = (e = this.options).onStateChange) == null || i.call(e, t));
   }
-  startWebSpeech(e) {
-    const i = new e();
-    i.continuous = !0, i.interimResults = !0, i.lang = this.options.language ?? "en-US", i.onstart = () => this.setListening(!0), i.onresult = (t) => {
-      var n, a, o, r, p;
+  startWebSpeech(t) {
+    const e = new t();
+    e.continuous = !0, e.interimResults = !0, e.lang = this.options.language ?? "en-US", e.onstart = () => this.setListening(!0), e.onresult = (i) => {
+      var n, o, h, c, f;
       let s = "";
-      for (let c = t.resultIndex; c < t.results.length; c += 1) {
-        const g = t.results[c], d = String(((n = g[0]) == null ? void 0 : n.transcript) ?? "").trim();
-        d && (g.isFinal ? (o = (a = this.options).onCommit) == null || o.call(a, d) : s += `${s ? " " : ""}${d}`);
+      for (let u = i.resultIndex; u < i.results.length; u += 1) {
+        const g = i.results[u], d = String(((n = g[0]) == null ? void 0 : n.transcript) ?? "").trim();
+        d && (g.isFinal ? (h = (o = this.options).onCommit) == null || h.call(o, d) : s += `${s ? " " : ""}${d}`);
       }
-      s && ((p = (r = this.options).onPartial) == null || p.call(r, s));
-    }, i.onerror = (t) => {
-      var n, a;
-      const s = t == null ? void 0 : t.error;
-      s === "no-speech" || s === "aborted" || (this.stopRequested = !0, (a = (n = this.options).onError) == null || a.call(n, new Error(`Speech recognition error: ${s}`)));
-    }, i.onend = () => {
-      if (this.stopRequested || this.recognition !== i) {
+      s && ((f = (c = this.options).onPartial) == null || f.call(c, s));
+    }, e.onerror = (i) => {
+      var n, o;
+      const s = i == null ? void 0 : i.error;
+      s === "no-speech" || s === "aborted" || (this.stopRequested = !0, (o = (n = this.options).onError) == null || o.call(n, new Error(`Speech recognition error: ${s}`)));
+    }, e.onend = () => {
+      if (this.stopRequested || this.recognition !== e) {
         this.recognition = null, this.setListening(!1);
         return;
       }
       try {
-        i.start();
+        e.start();
       } catch {
         this.recognition = null, this.setListening(!1);
       }
-    }, this.recognition = i, i.start();
+    }, this.recognition = e, e.start();
   }
   async startMoonshine() {
-    const e = await import("@moonshine-ai/moonshine-js"), i = new e.MicrophoneTranscriber(
+    const t = await import("@moonshine-ai/moonshine-js"), e = new t.MicrophoneTranscriber(
       this.options.model ?? "model/small",
       {
-        onTranscriptionUpdated: (t) => {
+        onTranscriptionUpdated: (i) => {
           var s, n;
-          (n = (s = this.options).onPartial) == null || n.call(s, String(t ?? "").trim());
+          (n = (s = this.options).onPartial) == null || n.call(s, String(i ?? "").trim());
         },
-        onTranscriptionCommitted: (t) => {
-          var n, a, o, r;
-          const s = String(t ?? "").trim();
-          s ? (a = (n = this.options).onCommit) == null || a.call(n, s) : (r = (o = this.options).onPartial) == null || r.call(o, "");
+        onTranscriptionCommitted: (i) => {
+          var n, o, h, c;
+          const s = String(i ?? "").trim();
+          s ? (o = (n = this.options).onCommit) == null || o.call(n, s) : (c = (h = this.options).onPartial) == null || c.call(h, "");
         }
       },
       !1
       // streaming mode
     );
-    if (this.moonshine = i, await i.start(), this.stopRequested) {
+    if (this.moonshine = e, await e.start(), this.stopRequested) {
       await this.stop();
       return;
     }
@@ -294,8 +265,12 @@ class j {
   }
 }
 export {
-  j as LiveTranscriber,
-  A as ReadAloudController,
-  P as isTranscriptionSupported
+  M as LiveTranscriber,
+  O as ReadAloudController,
+  x as isTranscriptionSupported,
+  E as looksLikeMarkdown,
+  L as markdownToSpeech,
+  D as markdownToSpeechSegments,
+  N as stripInlineMarkdown
 };
 //# sourceMappingURL=client.js.map