Update app.py

app.py

@@ -27,7 +27,6 @@ def retry_on_rate_limit(func, max_retries=3, initial_delay=5, backoff=2):
             try:
                 return func(*args, **kwargs)
             except Exception as e:
-                # You can catch a more specific HF exception if desired
                 if attempt < max_retries - 1:
                     print(f"Rate limit or error, retrying in {delay}s…")
                     time.sleep(delay)
@@ -77,11 +76,7 @@ def extract_frame(video_path: str, timestamp: str, output_path: str) -> None:
 
 @retry_on_rate_limit
 def transcribe_audio(asr_pipeline, audio_path: str) -> list[dict]:
-    """
-    Uses Whisper via HF pipeline to get timestamped chunks.
-    """
     result = asr_pipeline(audio_path)
-    # Whisper pipeline returns `chunks` when return_timestamps is set
     return result.get("chunks", [{"text": result["text"], "timestamp": (0.0, 0.0)}])
 
 
@@ -106,13 +101,11 @@ def run_pipeline(hf_token: str, video_file: str) -> list[dict]:
         model="openai/whisper-large-v2",
         chunk_length_s=30,
         stride_length_s=(5, 5),
-        return_timestamps="sentence",
-        use_auth_token=hf_token
+        return_timestamps="sentence"
     )
     summarizer = pipeline(
         "summarization",
-        model="deepseek/deepseek-v2",
-        use_auth_token=hf_token
+        model="deepseek/deepseek-v2"
     )
 
     # —– 1. Chunk the video —–
@@ -126,8 +119,8 @@ def run_pipeline(hf_token: str, video_file: str) -> list[dict]:
         for c in transcribe_audio(asr, wav):
             segments.append({
                 "text": c["text"],
-                "start": f"{c['timestamp'][0]:02d}:{(c['timestamp'][0]%60):06.3f}",
-                "end":   f"{c['timestamp'][1]:02d}:{(c['timestamp'][1]%60):06.3f}"
+                "start": f"{int(c['timestamp'][0]//60):02d}:{c['timestamp'][0]%60:06.3f}",
+                "end":   f"{int(c['timestamp'][1]//60):02d}:{c['timestamp'][1]%60:06.3f}"
             })
 
     # —– 3. Break into blocks, summarize, and extract phrases —–
@@ -139,7 +132,7 @@ def run_pipeline(hf_token: str, video_file: str) -> list[dict]:
     Path("frames").mkdir(exist_ok=True)
     frames = []
     for seg in segments:
-        ts_clean = seg["start"].replace(":", "-").replace(".", "-")
+        ts_clean = seg["start"].replace(":", "-")
         out = f"frames/frame_{ts_clean}.jpg"
         extract_frame(video_file, seg["start"], out)
         frames.append(out)