Update app.py

app.py

@@ -1,15 +1,17 @@
 import gradio as gr
 import yt_dlp
 import os
+import shutil
 from faster_whisper import WhisperModel
 
-# --- 1. Model Setup ---
+# --- 1. Model Setup (Base Model = Sahi Accuracy) ---
 model = None
 
 def load_model():
     global model
     if model is None:
-        print("📥 Loading Whisper Model...")
+        print("📥 Loading Base Model (Best Balance)...")
+        # 'base' model accuracy aur speed ka perfect mix hai
         model = WhisperModel("base", device="cpu", compute_type="int8")
         print("✅ Model Loaded!")
     return model
@@ -17,26 +19,22 @@ def load_model():
 # --- 2. Process Audio ---
 def process_audio(url):
     if not url:
-        return "⚠️ कृपया URL डालें।"
-
-    print(f"Processing: {url}")
-    
-    # Filename Setup
+        return "⚠️ URL missing!"
+        
     output_audio = "tiktok_audio"
     if os.path.exists(f"{output_audio}.mp3"): 
         os.remove(f"{output_audio}.mp3")
 
-    # FIX: FFmpeg ka location folder (jahan ffmpeg aur ffprobe dono hote hain)
-    ffmpeg_dir = "/usr/bin"
+    ffmpeg_dir = "/usr/bin" # System Path
 
     ydl_opts = {
         'format': 'bestaudio/best',
         'outtmpl': output_audio,
-        'ffmpeg_location': ffmpeg_dir,  # <--- DIRECT PATH FIX
+        'ffmpeg_location': ffmpeg_dir,
         'postprocessors': [{
             'key': 'FFmpegExtractAudio',
             'preferredcodec': 'mp3',
-            'preferredquality': '192',
+            'preferredquality': '192', # Quality wapas badha di taaki shabd saaf sunayi dein
         }],
         'quiet': True,
         'no_warnings': True,
@@ -47,35 +45,46 @@ def process_audio(url):
         }
     }
 
-    # 1. Download
     try:
         with yt_dlp.YoutubeDL(ydl_opts) as ydl:
             ydl.download([url])
     except Exception as e:
-        return f"❌ Download Error: {str(e)}\n(Ensure packages.txt has 'ffmpeg')"
+        return f"❌ Download Error: {str(e)}"
 
-    # 2. Transcribe
     if not os.path.exists(f"{output_audio}.mp3"):
-        return "❌ Error: Audio file download nahi ho payi."
+        return "❌ Error: Audio file nahi mili."
 
+    # --- Transcribe (Smart Settings) ---
     try:
         current_model = load_model()
-        # Transcribe directly from mp3
-        segments, _ = current_model.transcribe(f"{output_audio}.mp3", beam_size=5)
+        
+        # 'beam_size=1' rakha hai taaki speed tez rahe
+        # Lekin model 'base' hai to accuracy acchi rahegi
+        segments, _ = current_model.transcribe(
+            f"{output_audio}.mp3", 
+            beam_size=1, 
+            vad_filter=True
+        )
+        
         text = " ".join([s.text for s in segments])
         return text.strip()
     except Exception as e:
         return f"Transcription Error: {str(e)}"
 
 # --- 3. UI ---
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 📝 Professional TikTok Transcriber")
+css = """
+.container {max-width: 800px; margin: auto;}
+.gr-button-primary {background-color: #2563eb !important; color: white !important;} 
+"""
+
+with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
+    gr.Markdown("# 📝 Accurate TikTok Transcriber")
     
     with gr.Row():
         link_input = gr.Textbox(label="TikTok URL", placeholder="Paste Link Here...")
         btn = gr.Button("Transcribe", variant="primary")
     
-    transcript_out = gr.Code(label="Transcript", language="markdown", interactive=False, lines=15)
+    transcript_out = gr.Code(label="Transcript Result", language="markdown", interactive=False, lines=15)
 
     btn.click(fn=process_audio, inputs=link_input, outputs=transcript_out)