11

scripts/nodejs/pipeline_native.js

@@ -264,6 +264,42 @@ function parseTranscriptionText(trans) {
   return fullText;
 }
 
+/** @returns {Array<{ text: string, start: number, end: number }>} start/end 毫秒 */
+function normalizeAsrTimeToMs(value) {
+  const n = Number(value);
+  if (!Number.isFinite(n) || n < 0) return 0;
+  // 对齐 Electron generateAssSubtitle：>100 视为毫秒，否则视为秒
+  if (n > 100) return Math.round(n);
+  return Math.round(n * 1000);
+}
+
+function normalizeAsrRecord(record) {
+  const text = String(record?.text || record?.sentence || "").trim();
+  if (!text) return null;
+  let start = normalizeAsrTimeToMs(record.start ?? record.begin ?? record.beginTime ?? 0);
+  let end = normalizeAsrTimeToMs(record.end ?? record.endTime ?? start + 2000);
+  if (end <= start) end = start + 2000;
+  return { text, start, end };
+}
+
+/** @returns {Array<{ text: string, start: number, end: number }>} */
+function normalizeAsrRecords(records) {
+  if (!Array.isArray(records)) return [];
+  const normalized = records.map(normalizeAsrRecord).filter(Boolean);
+  const MIN_GAP_MS = 50;
+  for (let i = 0; i < normalized.length - 1; i++) {
+    const current = normalized[i];
+    const next = normalized[i + 1];
+    if (current.end >= next.start) {
+      current.end = Math.max(current.start + MIN_GAP_MS, next.start - MIN_GAP_MS);
+    }
+    if (current.end <= current.start) {
+      current.end = current.start + MIN_GAP_MS;
+    }
+  }
+  return normalized;
+}
+
 /** @returns {Array<{ text: string, start: number, end: number }>} start/end 毫秒 */
 function parseTranscriptionSentences(trans) {
   const records = [];
@@ -273,14 +309,16 @@ function parseTranscriptionSentences(trans) {
         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);
+          const start = normalizeAsrTimeToMs(s.begin_time ?? s.start_time ?? 0);
+          const end = normalizeAsrTimeToMs(
+            s.end_time ?? s.end ?? start + 2000,
+          );
           records.push({ text, start, end });
         }
       }
     }
   }
-  return records;
+  return normalizeAsrRecords(records);
 }
 
 async function fetchJson(url, options = {}) {
@@ -624,15 +662,21 @@ async function localAsr(audioPath, info) {
                 data.records ||
                 [];
               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)
+                ? normalizeAsrRecords(
+                    records
+                      .map((r) => {
+                        const text = String(r.text || "").trim();
+                        if (!text) return null;
+                        const start = normalizeAsrTimeToMs(
+                          r.start ?? r.begin_time ?? 0,
+                        );
+                        const end = normalizeAsrTimeToMs(
+                          r.end ?? r.end_time ?? start + 2000,
+                        );
+                        return { text, start, end };
+                      })
+                      .filter(Boolean),
+                  )
                 : [];
               const text = sentences.length
                 ? sentences.map((r) => r.text).join("")