perf(tts): reduce free request count by merging non-speaker segments

modules/tts/tts-text.js

@@ -169,7 +169,7 @@ export function parseTtsSegments(text) {
 
 // ============ 非鉴权分段切割 ============
 
-const FREE_MAX_TEXT = 200;
+const FREE_MAX_TEXT = 1000;
 const FREE_MIN_TEXT = 50;
 const FREE_SENTENCE_DELIMS = new Set(['。', '！', '？', '!', '?', ';', '；', '…', '.', '，', ',', '、', ':', '：']);
 
@@ -230,8 +230,46 @@ function splitTextForFree(text, maxLength = FREE_MAX_TEXT) {
 
 export function splitTtsSegmentsForFree(segments, maxLength = FREE_MAX_TEXT) {
     if (!Array.isArray(segments) || !segments.length) return [];
-    const out = [];
+    const normalizedSegments = [];
+
+    // In free mode, only explicit speaker directives are semantic split points.
+    // Adjacent segments without speaker= are merged to reduce request count.
+    let mergeBuffer = null;
+    const flushMergeBuffer = () => {
+        if (!mergeBuffer) return;
+        normalizedSegments.push(mergeBuffer);
+        mergeBuffer = null;
+    };
+
     for (const seg of segments) {
+        const hasExplicitSpeaker = !!String(seg?.speaker || '').trim();
+        const text = String(seg?.text || '').trim();
+        if (!text) continue;
+
+        if (hasExplicitSpeaker) {
+            flushMergeBuffer();
+            normalizedSegments.push({
+                ...seg,
+                text,
+            });
+            continue;
+        }
+
+        if (!mergeBuffer) {
+            mergeBuffer = {
+                ...seg,
+                text,
+                speaker: '',
+            };
+            continue;
+        }
+
+        mergeBuffer.text += `\n${text}`;
+    }
+    flushMergeBuffer();
+
+    const out = [];
+    for (const seg of normalizedSegments) {
         const parts = splitTextForFree(seg.text, maxLength);
         if (!parts.length) continue;
         let buffer = '';