11

scripts/nodejs/pipeline_native.js

@@ -259,6 +259,25 @@ function parseTranscriptionText(trans) {
   return fullText;
 }
 
+/** @returns {Array<{ text: string, start: number, end: number }>} start/end 毫秒 */
+function parseTranscriptionSentences(trans) {
+  const records = [];
+  if (trans?.transcripts?.length) {
+    for (const t of trans.transcripts) {
+      if (t.sentences?.length) {
+        for (const s of t.sentences) {
+          const text = String(s.text || "").trim();
+          if (!text) continue;
+          const start = Number(s.begin_time ?? s.start_time ?? 0);
+          const end = Number(s.end_time ?? s.end_time ?? start + 2000);
+          records.push({ text, start, end });
+        }
+      }
+    }
+  }
+  return records;
+}
+
 async function fetchJson(url, options = {}) {
   const res = await fetch(url, options);
   const text = await res.text();
@@ -537,7 +556,8 @@ async function aliyunAsrFiletrans(audioUrl, opts) {
       }
       const trans = await fetchJson(transUrl);
       const fullText = parseTranscriptionText(trans);
-      return { success: true, text: fullText };
+      const sentences = parseTranscriptionSentences(trans);
+      return { success: true, text: fullText, sentences };
     }
     if (status === "FAILED" || status === "UNKNOWN") {
       const detail =
@@ -597,10 +617,21 @@ async function localAsr(audioPath, info) {
                 data.data?.records ||
                 data.records ||
                 [];
-              const text = Array.isArray(records)
-                ? records.map((r) => r.text || "").join("")
+              const sentences = Array.isArray(records)
+                ? records
+                    .map((r) => {
+                      const text = String(r.text || "").trim();
+                      if (!text) return null;
+                      const start = Number(r.start ?? r.begin_time ?? 0);
+                      const end = Number(r.end ?? r.end_time ?? start + 2000);
+                      return { text, start, end };
+                    })
+                    .filter(Boolean)
+                : [];
+              const text = sentences.length
+                ? sentences.map((r) => r.text).join("")
                 : data.text || data.data?.text || "";
-              resolve({ success: true, text: String(text) });
+              resolve({ success: true, text: String(text), sentences });
             } else {
               resolve({ success: false, error: data.msg || data.error || raw.slice(0, 200) });
             }
@@ -656,8 +687,10 @@ module.exports = {
   aliyunOssUpload,
   aliyunAsrFiletrans,
   localAsr,
+  parseTranscriptionSentences,
   openaiChat,
   resolveTtsModel,
   splitTextForTts,
   qwenTtsSynthesize,
+  locateFfmpeg,
 };