@ww_nero/media 1.3.1 → 1.3.2

package/index.js

@@ -114,7 +114,7 @@ const resolveAudioFile = (workingDir, rawPath) => {
 const server = new Server(
   {
     name: 'media',
-    version: '1.3.1',
+    version: '1.3.2',
   },
   {
     capabilities: {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ww_nero/media",
-  "version": "1.3.1",
+  "version": "1.3.2",
   "description": "MCP server for media processing, including ASR speech recognition and TTS speech synthesis",
   "main": "index.js",
   "bin": {

package/utils/asr.js

@@ -8,6 +8,8 @@ const ASR_SERVICE_URL = 'https://dashscope.aliyuncs.com/api/v1/services/audio/as
 const SUPPORTED_AUDIO_TYPES = ['.mp3', '.wav'];
 const MAX_FILE_SIZE_BYTES = 120 * 1024 * 1024; // 120MB
 const LANGUAGE_HINTS = ['zh', 'en', 'ja'];
+const SENTENCE_SPLIT_REGEX = /[\u3002\uff01\uff1f\uff0c\u3001\uff1b\uff1a.,!?;:]+/u;
+const PUNCTUATION_REMOVE_REGEX = /[\p{P}\p{S}]/gu;
 
 /**
  * 将毫秒转换为 SRT 时间格式 HH:MM:SS,mmm
@@ -20,8 +22,37 @@ const msToSrtTime = (ms) => {
   return `${String(hours).padStart(2, '0')}:${String(minutes).padStart(2, '0')}:${String(seconds).padStart(2, '0')},${String(milliseconds).padStart(3, '0')}`;
 };
 
+const normalizeClause = (text) => {
+  if (!text) return { text: '', charCount: 0 };
+
+  const withoutPunctuation = text
+    .replace(PUNCTUATION_REMOVE_REGEX, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+
+  const charCount = withoutPunctuation.replace(/\s+/g, '').length;
+  return { text: withoutPunctuation, charCount };
+};
+
+const splitSentence = (text) => {
+  if (!text || typeof text !== 'string') return [];
+  const rawClauses = text.split(SENTENCE_SPLIT_REGEX);
+  const clauses = [];
+
+  for (const clause of rawClauses) {
+    const { text: cleanedText, charCount } = normalizeClause(clause);
+    if (cleanedText && charCount > 0) {
+      clauses.push({ text: cleanedText, charCount });
+    }
+  }
+
+  return clauses;
+};
+
 /**
  * 将 ASR 识别结果转换为 SRT 字幕内容
+ * 1. 句子按标点拆分为子句，去除所有标点符号
+ * 2. 按子句字符数比例分配句子时间轴
  */
 const asrToSrt = (asrData) => {
   const srtEntries = [];
@@ -35,17 +66,34 @@ const asrToSrt = (asrData) => {
       const sentences = transcript.sentences || [];
 
       for (const sentence of sentences) {
-        const beginTime = sentence.begin_time || 0;
-        const endTime = sentence.end_time || 0;
-        const text = (sentence.text || '').trim();
+        const beginTime = Number.isFinite(sentence.begin_time) ? sentence.begin_time : 0;
+        const endTime = Number.isFinite(sentence.end_time) ? sentence.end_time : 0;
+        const duration = Math.max(endTime - beginTime, 0);
+        const clauses = splitSentence(sentence.text || '');
+
+        const totalChars = clauses.reduce((sum, c) => sum + c.charCount, 0);
+        if (!totalChars || !clauses.length) {
+          continue;
+        }
+
+        let accumulatedChars = 0;
 
-        if (text) {
-          const startStr = msToSrtTime(beginTime);
-          const endStr = msToSrtTime(endTime);
+        clauses.forEach((clause, idx) => {
+          const startMs = Math.round(
+            beginTime + (duration * accumulatedChars) / totalChars
+          );
+          accumulatedChars += clause.charCount;
+          const endMs =
+            idx === clauses.length - 1
+              ? endTime
+              : Math.round(beginTime + (duration * accumulatedChars) / totalChars);
 
-          srtEntries.push(`${subtitleIndex}\n${startStr} --> ${endStr}\n${text}\n`);
+          const startStr = msToSrtTime(startMs);
+          const endStr = msToSrtTime(endMs);
+
+          srtEntries.push(`${subtitleIndex}\n${startStr} --> ${endStr}\n${clause.text}\n`);
           subtitleIndex++;
-        }
+        });
       }
     }
   }