freezmo/lib/speech.ts

"use client";

/**
 * Text-to-speech for study cards, on the browser's own Web Speech API.
 *
 * No dependency and no audio leaves the app ourselves: the voice, the engine
 * and any cloud fallback belong to the operating system/browser. The module
 * keeps one utterance at a time app-wide (two cards reading over each other is
 * never what anyone wants) and notifies subscribers so the button that started
 * it can flip to "stop".
 */

const listeners = new Set<() => void>();
let speakingId: string | null = null;

function notify(): void {
  listeners.forEach((listener) => listener());
}

export function speechAvailable(): boolean {
  return typeof window !== "undefined" && "speechSynthesis" in window && typeof SpeechSynthesisUtterance !== "undefined";
}

/** Language names ("Czech") and codes ("cs") → a BCP-47 tag for the voice. */
const LOCALES: Record<string, string> = {
  english: "en-US",
  czech: "cs-CZ",
  slovak: "sk-SK",
  polish: "pl-PL",
  german: "de-DE",
  french: "fr-FR",
  spanish: "es-ES",
  italian: "it-IT",
  portuguese: "pt-PT",
  dutch: "nl-NL",
  swedish: "sv-SE",
  norwegian: "nb-NO",
  danish: "da-DK",
  finnish: "fi-FI",
  hungarian: "hu-HU",
  romanian: "ro-RO",
  turkish: "tr-TR",
  croatian: "hr-HR",
  slovenian: "sl-SI",
  serbian: "sr-RS",
  russian: "ru-RU",
  ukrainian: "uk-UA",
  bulgarian: "bg-BG",
  belarusian: "be-BY",
  macedonian: "mk-MK",
  greek: "el-GR",
  chinese: "zh-CN",
  mandarin: "zh-CN",
  japanese: "ja-JP",
  korean: "ko-KR",
  arabic: "ar-SA",
  persian: "fa-IR",
  farsi: "fa-IR",
  urdu: "ur-PK",
  hebrew: "he-IL",
  hindi: "hi-IN",
  marathi: "mr-IN",
  nepali: "ne-NP",
  thai: "th-TH",
  vietnamese: "vi-VN",
};

/**
 * A BCP-47 tag for the deck's language, falling back to the interface locale
 * (so an untagged deck reads with the voice the user already hears the app in).
 */
export function speechLocale(language: string | undefined, fallback: string): string {
  const key = (language ?? "").trim().toLowerCase().replace(/[_.]/g, "-").split("-")[0].replace(/[^a-z]/g, "");
  if (key && LOCALES[key]) return LOCALES[key];
  if (key.length === 2) return language!.trim();
  return fallback === "cs" ? "cs-CZ" : "en-US";
}

/** The installed voice closest to `locale` (exact match, then language).
 *  Shared with the podcast player (lib/podcast-speech). */
export function pickVoice(locale: string): SpeechSynthesisVoice | null {
  try {
    const voices = window.speechSynthesis.getVoices();
    if (voices.length === 0) return null;
    const wanted = locale.toLowerCase();
    const prefix = wanted.split("-")[0];
    return (
      voices.find((voice) => voice.lang.toLowerCase() === wanted) ??
      voices.find((voice) => voice.lang.toLowerCase().replace("_", "-") === wanted) ??
      voices.find((voice) => voice.lang.toLowerCase().startsWith(`${prefix}-`) || voice.lang.toLowerCase() === prefix) ??
      null
    );
  } catch {
    return null;
  }
}

/**
 * Reads `text` aloud under `id` (usually the button's own React id, so the
 * button that started it is the one that shows "stop").
 */
export function speak(id: string, text: string, locale: string): void {
  if (!speechAvailable()) return;
  const clean = text.replace(/…/g, ", ").replace(/_{2,}/g, "… ").trim();
  if (!clean) return;
  try {
    window.speechSynthesis.cancel();
    const utterance = new SpeechSynthesisUtterance(clean);
    utterance.lang = locale;
    utterance.rate = 0.95;
    const voice = pickVoice(locale);
    if (voice) utterance.voice = voice;
    const done = () => {
      if (speakingId === id) {
        speakingId = null;
        notify();
      }
    };
    utterance.onend = done;
    utterance.onerror = done;
    speakingId = id;
    notify();
    window.speechSynthesis.speak(utterance);
  } catch {
    speakingId = null;
    notify();
  }
}

export function stopSpeaking(): void {
  if (!speechAvailable()) return;
  try {
    window.speechSynthesis.cancel();
  } catch {
    /* already gone */
  }
  if (speakingId !== null) {
    speakingId = null;
    notify();
  }
}

/** True while `id`'s utterance is the one playing. */
export function isSpeaking(id: string): boolean {
  return speakingId === id;
}

export function subscribeSpeech(listener: () => void): () => void {
  listeners.add(listener);
  return () => listeners.delete(listener);
}