/** * The browser's speech engine reads one utterance at a time. A single long * utterance stops early in desktop Chrome, and line breaks disappear, so the * scene becomes sentence-sized chunks and every block stays separate. */ import { sanitiseBody } from '#lib/body.ts'; export const MAX_SPEECH_CHARS = 140; export interface SpeechEngine { readonly speaking: boolean; readonly paused: boolean; speak(utterance: SpeechSynthesisUtterance): void; cancel(): void; pause(): void; resume(): void; } type DocNode = { type?: string; text?: string; content?: DocNode[]; }; /** * `Intl.Segmenter` splits after `Mr.` and after an initial, so those pieces are * joined again. A wrong split adds an audible pause in the middle of a name; a * wrong join only makes one utterance longer. */ const ABBREVIATION = /(?:\b(?:Mr|Mrs|Ms|Dr|Prof|Sr|Jr|St|vs|etc|No|Vol|Ch|Fig|Ave|Rd|Mt|Ft)\.$|^(?:[A-Z]\.\s*)+$|(?:\b[A-Z]\.){2,}$|e\.g\.$|i\.e\.$)/; export function speechChunks(body: unknown, maxChars = MAX_SPEECH_CHARS): string[] { const limit = Math.max(1, Math.floor(maxChars)); const chunks: string[] = []; for (const block of blocks(body)) { for (const sentence of sentences(block)) { for (const piece of splitLong(sentence, limit)) { const text = piece.trim(); if (text.length > 0) chunks.push(text); } } } return chunks; } export function speechSupported(): boolean { return ( typeof window !== 'undefined' && typeof window.speechSynthesis === 'object' && window.speechSynthesis !== null && typeof SpeechSynthesisUtterance !== 'undefined' ); } export function browserSpeechEngine(): SpeechEngine | null { return speechSupported() ? window.speechSynthesis : null; } export function preferredLanguage(): string { if (typeof document !== 'undefined' && document.documentElement.lang.length > 0) { return document.documentElement.lang; } if (typeof navigator !== 'undefined' && navigator.language.length > 0) return navigator.language; return 'en'; } export function sameLanguage(a: string, b: string): boolean { const primary = (tag: string) => tag.toLowerCase().split(/[-_]/)[0] ?? ''; return primary(a).length > 0 && primary(a) === primary(b); } /** A voice that stays on the device and matches the page's language. */ export function localVoiceFor( voices: readonly SpeechSynthesisVoice[], language: string ): SpeechSynthesisVoice | null { const local = voices.filter((voice) => voice.localService && sameLanguage(voice.lang, language)); return local.find((voice) => voice.default) ?? local[0] ?? null; } function blocks(body: unknown): string[] { const doc = sanitiseBody(body); if (!doc) return []; const found: string[] = []; for (const child of doc.content) walk(child, found); return found; } function walk(node: DocNode, found: string[]) { switch (node.type) { case 'paragraph': case 'heading': case 'codeBlock': case 'text': pushText(node, found); return; case 'horizontalRule': return; default: if (Array.isArray(node.content) && node.content.some(isBlock)) { for (const child of node.content) walk(child, found); } else { pushText(node, found); } } } function isBlock(node: DocNode): boolean { return ( node.type === 'paragraph' || node.type === 'heading' || node.type === 'blockquote' || node.type === 'bulletList' || node.type === 'orderedList' || node.type === 'listItem' || node.type === 'codeBlock' || node.type === 'horizontalRule' ); } function pushText(node: DocNode, found: string[]) { const text = textOf(node).replace(/\s+/g, ' ').trim(); if (text.length > 0) found.push(text); } function textOf(node: DocNode): string { if (typeof node.text === 'string') return node.text; if (node.type === 'hardBreak') return ' '; if (!Array.isArray(node.content)) return ''; return node.content.map(textOf).join(''); } function sentences(text: string): string[] { const parts = segment(text); const merged: string[] = []; for (const part of parts) { const sentence = part.trim(); if (sentence.length === 0) continue; const previous = merged.at(-1); if (previous && ABBREVIATION.test(previous)) { merged[merged.length - 1] = `${previous} ${sentence}`; } else { merged.push(sentence); } } return merged; } function segment(text: string): string[] { if (typeof Intl !== 'undefined' && typeof Intl.Segmenter === 'function') { const segmenter = new Intl.Segmenter(undefined, { granularity: 'sentence' }); return Array.from(segmenter.segment(text), (part) => part.segment); } return text.match(/[^.!?…]+(?:[.!?…]+["'”’)\]]*|$)/g) ?? [text]; } function splitLong(text: string, maxChars: number): string[] { const pieces: string[] = []; let rest = text; while (rest.length > maxChars) { let cut = rest.lastIndexOf(' ', maxChars); if (cut <= 0) cut = maxChars; // Keep a surrogate pair together; a split inside one is not a character. if (isHighSurrogate(rest[cut - 1]) && isLowSurrogate(rest[cut])) cut -= 1; if (cut <= 0) cut = Math.min(2, rest.length); const piece = rest.slice(0, cut).trim(); if (piece.length > 0) pieces.push(piece); rest = rest.slice(cut).trim(); } if (rest.length > 0) pieces.push(rest); return pieces; } function isHighSurrogate(value: string | undefined): boolean { if (!value) return false; const code = value.charCodeAt(0); return code >= 0xd800 && code <= 0xdbff; } function isLowSurrogate(value: string | undefined): boolean { if (!value) return false; const code = value.charCodeAt(0); return code >= 0xdc00 && code <= 0xdfff; }