diff --git a/speech.js b/speech.js
index 2a4232a..d6f11cb 100644
--- a/speech.js
+++ b/speech.js
@@ -1,5 +1,6 @@
 var speechVoices = [];
 var speechVoicesLoaded = false;
+var speechTimer = null;
 
 function refreshSpeechVoices() {
   if (!window.speechSynthesis) return;
@@ -36,28 +37,20 @@ function buildSpeechUtterance(text, lang) {
 }
 
 function doSpeak(text, lang) {
-  if (!window.speechSynthesis || !text) return;
+  if (!window.speechSynthesis || !window.SpeechSynthesisUtterance || !text) return;
   window.speechSynthesis.cancel();
-  window.speechSynthesis.speak(buildSpeechUtterance(text, lang));
+  if (speechTimer) clearTimeout(speechTimer);
+  var utterance = buildSpeechUtterance(text, lang);
+  speechTimer = setTimeout(function() {
+    window.speechSynthesis.speak(utterance);
+    if (window.speechSynthesis.paused) window.speechSynthesis.resume();
+  }, 60);
 }
 
 function speakText(text, lang) {
   if (!text || !window.speechSynthesis) return;
   refreshSpeechVoices();
-  if (speechVoicesLoaded) {
-    doSpeak(text, lang);
-    return;
-  }
-
-  var attempts = 0;
-  var timer = setInterval(function() {
-    attempts++;
-    refreshSpeechVoices();
-    if (speechVoicesLoaded || attempts >= 8) {
-      clearInterval(timer);
-      doSpeak(text, lang);
-    }
-  }, 80);
+  doSpeak(text, lang);
   window.speechSynthesis.getVoices();
 }
 
@@ -71,3 +64,12 @@ function initSpeech() {
 
 if (document.readyState === 'complete') initSpeech();
 else document.addEventListener('DOMContentLoaded', initSpeech);
+
+function warmSpeechVoices() {
+  if (!window.speechSynthesis) return;
+  window.speechSynthesis.getVoices();
+  refreshSpeechVoices();
+}
+
+document.addEventListener('touchstart', warmSpeechVoices, { once: true, passive: true });
+document.addEventListener('pointerdown', warmSpeechVoices, { once: true, passive: true });