Fix TTS handler to handle longer texts by breaking them into sentence queues

This commit is contained in:
2025-04-01 10:53:27 +02:00
parent bf62c63198
commit 113e3b995d
+215 -51
View File
@@ -13,6 +13,10 @@ class TTSHandler {
this.preferredVoice = null;
this.audioCache = new Map(); // Cache for audio segments
this.currentSpeed = 1.0;
this.hasUserActivation = false;
this.permissionError = false;
this.speakQueue = [];
this.isSpeakingFromQueue = false;
// Initialize if speech synthesis is available
if ('speechSynthesis' in window) {
@@ -27,10 +31,25 @@ class TTSHandler {
this.synth.onvoiceschanged = () => {
this.voiceCache = this.synth.getVoices();
this.selectPreferredVoice();
console.log("Voices loaded:", this.voiceCache.length);
};
// Enable by default if available
this.enabled = true;
// Disabled by default until user activates it
this.enabled = false;
// Set up periodic check to detect and fix stuck speech
setInterval(() => {
// If we think we're speaking but the browser doesn't, reset state
if (this.speaking && !this.synth.speaking && !this.isSpeakingFromQueue) {
console.log("Detected stuck speech state, resetting");
this.speaking = false;
// Try to continue the queue if there are more items
if (this.speakQueue.length > 0) {
this.processSpeakQueue();
}
}
}, 1000);
} else {
console.warn("Text-to-speech functionality not available in this browser.");
}
@@ -49,11 +68,15 @@ class TTSHandler {
"Karen"
];
// Debug: Print all available voices
console.log("Available voices:", this.voiceCache.map(v => v.name + " (" + v.lang + ")").join(", "));
// Try to find one of our preferred voices
for (const name of preferredVoiceNames) {
const voice = this.voiceCache.find(v => v.name === name);
if (voice) {
this.preferredVoice = voice;
console.log("Selected preferred voice:", name);
return;
}
}
@@ -62,28 +85,68 @@ class TTSHandler {
const englishVoice = this.voiceCache.find(v => v.lang.startsWith('en'));
if (englishVoice) {
this.preferredVoice = englishVoice;
console.log("Selected English voice:", englishVoice.name);
return;
}
// Last resort: use the first available voice
if (this.voiceCache.length > 0) {
this.preferredVoice = this.voiceCache[0];
console.log("Selected fallback voice:", this.voiceCache[0].name);
}
}
/**
* Toggle TTS functionality on/off
* @returns {boolean} New state of TTS (enabled/disabled)
*/
toggle() {
if (!this.synth) return false;
// Set user activation flag when toggle is called
this.hasUserActivation = true;
// Clear permission error on toggle
this.permissionError = false;
this.enabled = !this.enabled;
console.log("TTS toggled:", this.enabled ? "ON" : "OFF");
// Stop any ongoing speech when disabling
if (!this.enabled && this.speaking) {
this.stop();
}
// Try a test utterance to request permissions
if (this.enabled) {
try {
// Reset any current utterance first
this.synth.cancel();
this.speakQueue = [];
this.isSpeakingFromQueue = false;
// Create a silent utterance to trigger permission request
const testUtterance = new SpeechSynthesisUtterance("Hello");
testUtterance.volume = 0.05; // Very quiet but not silent to ensure it works
testUtterance.rate = 1.0;
// Handle any errors that might occur
testUtterance.onerror = (event) => {
console.warn("Permission error for TTS:", event);
if (event.error === "not-allowed") {
this.permissionError = true;
this.enabled = false;
alert("Text-to-speech was blocked by your browser. Please allow speech in your browser settings.");
}
};
// Try to speak the test utterance
this.synth.speak(testUtterance);
} catch (e) {
console.error("Failed to initialize TTS:", e);
}
}
return this.enabled;
}
@@ -93,11 +156,6 @@ class TTSHandler {
*/
setSpeed(speed) {
this.currentSpeed = Math.max(0.1, Math.min(2.0, speed));
if (this.utterance && this.speaking) {
// Cannot change speed of active utterance, need to restart
this.stop();
// Would need to restart the current text, but challenging without storing current text
}
}
/**
@@ -106,6 +164,8 @@ class TTSHandler {
* @returns {string} - Processed text
*/
processTextForSpeech(text) {
if (!text) return "";
// Remove markdown/formatting that would sound strange when read
text = text.replace(/\*\*([^*]+)\*\*/g, '$1'); // Bold
text = text.replace(/\*([^*]+)\*/g, '$1'); // Italic
@@ -118,66 +178,168 @@ class TTSHandler {
}
/**
* Add natural pauses after sentences using native TTS methods
* @param {SpeechSynthesisUtterance} utterance - The utterance to modify
* Split text into sentences for better speech handling
* @param {string} text - Text to split
* @returns {string[]} - Array of sentences
*/
addPauses(utterance) {
// Instead of modifying the text, we'll use the utterance's parameters
// to create natural pauses - these settings generally improve natural speaking
utterance.pitch = 1.0;
utterance.rate = this.currentSpeed;
splitIntoSentences(text) {
if (!text) return [];
// Some TTS engines support these parameters
if ('volume' in utterance) utterance.volume = 1.0;
// Split by sentence terminators, keeping the terminator with the sentence
const sentenceRegex = /[^.!?]+[.!?]+/g;
const sentences = text.match(sentenceRegex) || [text];
// If we have very long sentences, break them up by commas too
return sentences.reduce((result, sentence) => {
if (sentence.length > 150 && sentence.includes(',')) {
// Split long sentences at commas
const parts = sentence.split(/,\s*/);
for (let i = 0; i < parts.length - 1; i++) {
result.push(parts[i] + ',');
}
result.push(parts[parts.length - 1]);
return result;
}
result.push(sentence);
return result;
}, []);
}
/**
* Speak the provided text
* Speak a single utterance with proper configuration
* @param {string} text - Text to speak
* @param {function} onEndCallback - Callback to execute when finished
* @private
*/
speakUtterance(text, onEndCallback) {
if (!text || text.trim() === '') {
if (onEndCallback) onEndCallback();
this.processSpeakQueue();
return;
}
try {
const utterance = new SpeechSynthesisUtterance(text);
if (this.preferredVoice) {
utterance.voice = this.preferredVoice;
console.log("Using voice:", this.preferredVoice.name);
}
utterance.rate = this.currentSpeed;
utterance.pitch = 1.0;
utterance.volume = 1.0;
utterance.onstart = () => {
this.speaking = true;
console.log("TTS started speaking:", text.substring(0, 30) + "...");
};
utterance.onend = () => {
console.log("TTS finished speaking utterance");
if (onEndCallback) onEndCallback();
this.processSpeakQueue();
};
utterance.onerror = (event) => {
console.error("Speech synthesis error:", event);
if (event.error === "not-allowed") {
this.permissionError = true;
this.enabled = false;
}
if (onEndCallback) onEndCallback();
this.processSpeakQueue();
};
// Actually speak
this.synth.speak(utterance);
// Workaround for Chrome bug where speech synthesis gets stuck
if (!this.synth.speaking) {
this.synth.pause();
this.synth.resume();
}
} catch (e) {
console.error("Error in speakUtterance:", e);
if (onEndCallback) onEndCallback();
this.processSpeakQueue();
}
}
/**
* Process the next item in the speak queue
* @private
*/
processSpeakQueue() {
if (this.speakQueue.length === 0) {
this.isSpeakingFromQueue = false;
this.speaking = false;
return;
}
// Skip processing if we're already speaking (prevent overlapping sentences)
if (this.synth.speaking) {
setTimeout(() => this.processSpeakQueue(), 100);
return;
}
this.isSpeakingFromQueue = true;
const queueItem = this.speakQueue.shift();
console.log(`Speaking queue item (${this.speakQueue.length} remaining):`, queueItem.text.substring(0, 30) + "...");
this.speakUtterance(queueItem.text, queueItem.callback);
}
/**
* Speak the provided text by queueing sentences
* @param {string} text - Text to be spoken
* @param {function} onEndCallback - Callback when speech ends
* @param {function} onEndCallback - Callback when all speech ends
*/
speak(text, onEndCallback = null) {
if (!this.synth || !this.enabled || !text) return;
if (!this.synth || !this.enabled || !text) {
if (onEndCallback) onEndCallback();
return;
}
// Stop any existing speech
if (this.speaking) {
this.stop();
// Don't attempt to speak if there's been a permission error
if (this.permissionError) {
console.warn("Not attempting to speak due to permission error");
if (onEndCallback) onEndCallback();
return;
}
// Don't attempt to speak without user activation
if (!this.hasUserActivation) {
console.warn("Not attempting to speak because there hasn't been user interaction yet");
if (onEndCallback) onEndCallback();
return;
}
// Process text for better speech
const processedText = this.processTextForSpeech(text);
console.log("TTS attempting to speak:", processedText.substring(0, 50) + "...");
// Create and configure the utterance
this.utterance = new SpeechSynthesisUtterance(processedText);
// Stop any existing speech
this.stop();
if (this.preferredVoice) {
this.utterance.voice = this.preferredVoice;
// Split into sentences for better handling
const sentences = this.splitIntoSentences(processedText);
// Last sentence gets the callback
for (let i = 0; i < sentences.length; i++) {
this.speakQueue.push({
text: sentences[i],
callback: i === sentences.length - 1 ? onEndCallback : null
});
}
this.utterance.rate = this.currentSpeed;
this.utterance.pitch = 1.0;
// Apply natural pausing
this.addPauses(this.utterance);
// Set up event handlers
this.utterance.onstart = () => {
this.speaking = true;
};
this.utterance.onend = () => {
this.speaking = false;
if (onEndCallback) onEndCallback();
};
this.utterance.onerror = (event) => {
console.error("Speech synthesis error:", event);
this.speaking = false;
if (onEndCallback) onEndCallback();
};
// Start speaking
this.synth.speak(this.utterance);
// Start processing the queue if not already processing
if (!this.isSpeakingFromQueue) {
this.processSpeakQueue();
}
}
/**
@@ -210,13 +372,15 @@ class TTSHandler {
this.speaking = false;
this.paused = false;
this.utterance = null;
this.speakQueue = [];
this.isSpeakingFromQueue = false;
}
/**
* Check if TTS is currently active/enabled
*/
isEnabled() {
return this.enabled;
return this.enabled && !this.permissionError;
}
/**