Update app.py

app.py

@@ -20,6 +20,41 @@ def format_timestamp(seconds: float) -> str:
     millisecs = int((seconds - int(seconds)) * 1000)
     return f"{hours:02d}:{minutes:02d}:{secs:02d},{millisecs:03d}"
 
+def group_by_chars(result, max_chars: int):
+    """用词级时间戳按字数重新组合字幕段，时间戳精确来自 Whisper"""
+    words = []
+    for seg in result["segments"]:
+        for w in seg.get("words", []):
+            word = w["word"].strip()
+            if word:
+                words.append({"word": word, "start": w["start"], "end": w["end"]})
+
+    if not words:
+        # 没有词级时间戳，回退原始分段
+        return result["segments"]
+
+    segments = []
+    buf_text = ""
+    buf_start = None
+    buf_end = None
+
+    for w in words:
+        if buf_start is None:
+            buf_start = w["start"]
+        buf_text += w["word"]
+        buf_end = w["end"]
+
+        # 达到字数上限，或遇到句末标点就切断
+        if len(buf_text) >= max_chars or buf_text[-1] in "，。！？、,.!?":
+            segments.append({"start": buf_start, "end": buf_end, "text": buf_text})
+            buf_text = ""
+            buf_start = None
+
+    if buf_text:
+        segments.append({"start": buf_start, "end": buf_end, "text": buf_text})
+
+    return segments
+
 def generate_srt(segments) -> str:
     srt_content = []
     for i, segment in enumerate(segments, 1):
@@ -54,7 +89,8 @@ async def health_check():
 async def generate_subtitle(
     file: UploadFile = File(...),
     output_format: str = Form("srt"),
-    language: Optional[str] = Form(None)
+    language: Optional[str] = Form(None),
+    max_chars: int = Form(0)  # 0 = 不拆分，用 Whisper 原始分段
 ):
     valid_formats = ["srt", "vtt", "txt", "json"]
     if output_format not in valid_formats:
@@ -67,18 +103,20 @@ async def generate_subtitle(
         tmp_path = tmp_file.name
 
     try:
-        transcribe_options = {}
+        transcribe_options = {"word_timestamps": True}
         if language:
             transcribe_options["language"] = language
 
         result = model.transcribe(tmp_path, **transcribe_options)
 
+        segments = group_by_chars(result, max_chars) if max_chars > 0 else result["segments"]
+
         if output_format == "srt":
-            subtitle_text = generate_srt(result["segments"])
+            subtitle_text = generate_srt(segments)
             media_type = "text/plain"
             filename = "subtitle.srt"
         elif output_format == "vtt":
-            subtitle_text = generate_vtt(result["segments"])
+            subtitle_text = generate_vtt(segments)
             media_type = "text/plain"
             filename = "subtitle.vtt"
         elif output_format == "txt":
@@ -88,7 +126,7 @@ async def generate_subtitle(
         elif output_format == "json":
             subtitle_text = json.dumps({
                 "text": result["text"],
-                "segments": result["segments"],
+                "segments": segments,
                 "language": result["language"]
             }, ensure_ascii=False, indent=2)
             media_type = "application/json"