Update app.py

app.py

@@ -3,15 +3,22 @@ from fastapi.responses import PlainTextResponse
 import whisper
 import tempfile
 import os
+import sys
 import json
+import time
 from pathlib import Path
 from typing import Optional
 
+# 把 stderr 合并到 stdout，HuggingFace Spaces log 才能显示 Whisper 进度条
+sys.stderr = sys.stdout
+
 app = FastAPI(title="Subtitle Generator API")
 
-# ↓↓↓ 就改了这里：base → medium
 model_size = os.getenv("WHISPER_MODEL", "medium")
+print(f"[Startup] 加载 Whisper 模型: {model_size} ...", flush=True)
 model = whisper.load_model(model_size)
+print(f"[Startup] 模型加载完成 ✅", flush=True)
+
 
 def format_timestamp(seconds: float) -> str:
     hours = int(seconds // 3600)
@@ -20,6 +27,7 @@ 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 = []
@@ -30,7 +38,7 @@ def group_by_chars(result, max_chars: int):
                 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
+        print("[Warning] 没有词级时间戳，回退到原始分段", flush=True)
         return result["segments"]
 
     segments = []
@@ -44,7 +52,6 @@ def group_by_chars(result, max_chars: int):
         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 = ""
@@ -55,6 +62,7 @@ def group_by_chars(result, max_chars: int):
 
     return segments
 
+
 def generate_srt(segments) -> str:
     srt_content = []
     for i, segment in enumerate(segments, 1):
@@ -64,6 +72,7 @@ def generate_srt(segments) -> str:
         srt_content.append(f"{i}\n{start_time} --> {end_time}\n{text}\n")
     return "\n".join(srt_content)
 
+
 def generate_vtt(segments) -> str:
     vtt_content = ["WEBVTT\n"]
     for i, segment in enumerate(segments):
@@ -73,6 +82,7 @@ def generate_vtt(segments) -> str:
         vtt_content.append(f"{start_time} --> {end_time}\n{text}\n")
     return "\n".join(vtt_content)
 
+
 @app.get("/")
 async def root():
     return {
@@ -81,10 +91,12 @@ async def root():
         "device": "cpu"
     }
 
+
 @app.get("/health")
 async def health_check():
     return {"status": "healthy", "model": model_size}
 
+
 @app.post("/generate_subtitle")
 async def generate_subtitle(
     file: UploadFile = File(...),
@@ -102,14 +114,29 @@ async def generate_subtitle(
         tmp_file.write(content)
         tmp_path = tmp_file.name
 
+    file_mb = len(content) / 1024 / 1024
+    print(f"[Transcribe] 文件: {file.filename} ({file_mb:.1f} MB)", flush=True)
+    print(f"[Transcribe] 格式: {output_format} | 语言: {language or '自动检测'} | max_chars: {max_chars or '不拆分'}", flush=True)
+
     try:
-        transcribe_options = {"word_timestamps": True}
+        transcribe_options = {"word_timestamps": True, "verbose": True}
         if language:
             transcribe_options["language"] = language
 
+        print(f"[Transcribe] 开始转写，请稍候...", flush=True)
+        t0 = time.time()
         result = model.transcribe(tmp_path, **transcribe_options)
+        elapsed = time.time() - t0
 
-        segments = group_by_chars(result, max_chars) if max_chars > 0 else result["segments"]
+        seg_count = len(result["segments"])
+        detected_lang = result.get("language", "unknown")
+        print(f"[Transcribe] 完成 ✅ 耗时 {elapsed:.1f}s | 语言: {detected_lang} | 原始段数: {seg_count}", flush=True)
+
+        if max_chars > 0:
+            segments = group_by_chars(result, max_chars)
+            print(f"[Transcribe] 按字数拆分完成，共 {len(segments)} 段（max_chars={max_chars}）", flush=True)
+        else:
+            segments = result["segments"]
 
         if output_format == "srt":
             subtitle_text = generate_srt(segments)
@@ -132,6 +159,7 @@ async def generate_subtitle(
             media_type = "application/json"
             filename = "subtitle.json"
 
+        print(f"[Transcribe] 返回结果: {filename}", flush=True)
         return PlainTextResponse(
             content=subtitle_text,
             media_type=media_type,
@@ -139,11 +167,13 @@ async def generate_subtitle(
         )
 
     except Exception as e:
+        print(f"[Error] 转写失败: {e}", flush=True)
         raise HTTPException(status_code=500, detail=f"Transcription failed: {str(e)}")
     finally:
         if os.path.exists(tmp_path):
             os.unlink(tmp_path)
 
+
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run(app, host="0.0.0.0", port=7860)