Update app.py

app.py

@@ -49,7 +49,7 @@ def transcribe_audio(path):
     if not WHISPER_AVAILABLE:
         return "❌ Whisper not available. Please install openai-whisper."
     try:
-        model = whisper.load_model("tiny")
+        model = whisper.load_model("tiny.en")
         result = model.transcribe(path)
         return result["text"]
     except Exception as e:
@@ -109,6 +109,24 @@ def save_cookies(file):
         print(f"❌ Failed to handle cookies.txt: {e}")
         return None
 
+
+# ✅ Trim audio to shorter length (2 minutes) for CPU speed
+import subprocess
+
+def trim_audio(input_path, output_path, duration_sec=120):
+    try:
+        command = [
+            "ffmpeg", "-y", "-i", input_path,
+            "-t", str(duration_sec),  # duration in seconds
+            "-c", "copy", output_path
+        ]
+        subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        return output_path
+    except Exception as e:
+        print("❌ Error trimming audio:", e)
+        return input_path
+
+
 # YouTube flow
 def run_pipeline(url, cookies_file, show_transcript):
     try:
@@ -122,7 +140,11 @@ def run_pipeline(url, cookies_file, show_transcript):
         if not audio_path:
             return status, ""
 
-        transcript = transcribe_audio(audio_path)
+        # ⏱ Trim audio to 2 minutes before transcription
+        trimmed_path = tempfile.mktemp(suffix=".mp3")
+        trim_audio(audio_path, trimmed_path)
+
+        transcript = transcribe_audio(trimmed_path)
         if transcript.startswith("❌"):
             return transcript, ""
 
@@ -137,6 +159,7 @@ def run_pipeline(url, cookies_file, show_transcript):
         print(tb)
         return f"❌ Unhandled Error:\n{tb}", ""
 
+
 # Audio upload flow
 def run_pipeline_audio(audio_file, show_transcript):
     try:
@@ -145,6 +168,7 @@ def run_pipeline_audio(audio_file, show_transcript):
         if audio_file is None:
             return "❌ No audio file uploaded", ""
 
+        # Save uploaded file
         temp_audio_path = tempfile.mktemp(suffix=os.path.splitext(str(audio_file))[-1])
         if hasattr(audio_file, "read"):
             with open(temp_audio_path, "wb") as f:
@@ -152,7 +176,11 @@ def run_pipeline_audio(audio_file, show_transcript):
         else:
             shutil.copy(str(audio_file), temp_audio_path)
 
-        transcript = transcribe_audio(temp_audio_path)
+        # ⏱ Trim audio to 2 minutes
+        trimmed_path = tempfile.mktemp(suffix=".mp3")
+        trim_audio(temp_audio_path, trimmed_path)
+
+        transcript = transcribe_audio(trimmed_path)
         if transcript.startswith("❌"):
             return transcript, ""
 
@@ -167,6 +195,7 @@ def run_pipeline_audio(audio_file, show_transcript):
         print(tb)
         return f"❌ Unhandled Error:\n{tb}", ""
 
+
 # Gradio UI
 with gr.Blocks(title="Stock Insights from YouTube or Audio") as demo:
     gr.Markdown("""