Update app.py

app.py

@@ -79,31 +79,27 @@ class YouTubeDownloader:
                - Visual effects, text overlays, or graphics
                - Mood, tone, and atmosphere
                - Camera movements or angles (if apparent)
+
+            2. Dialogue Emphasis:
+               - Include short dialogue lines in **every scene** wherever plausible.
+               - Write lines like: Character: "Actual or inferred line..."
+               - If dialogue is not available, intelligently infer probable phrases (e.g., "Welcome!", "Try this now!", "It feels amazing!").
+               - Do NOT skip dialogue unless it’s clearly impossible.
             
-            2. Timestamp Guidelines:
+            3. Timestamp Guidelines:
                - For videos under 1 minute: 2-3 second segments
                - For videos 1-5 minutes: 3-5 second segments  
                - For videos 5-15 minutes: 5-10 second segments
                - For videos over 15 minutes: 10-15 second segments
                - Maximum 20 scenes total for longer videos
             
-            3. Format each scene EXACTLY like this:
+            4. Format each scene EXACTLY like this:
                **[MM:SS-MM:SS]**: Detailed description including who is visible, what they're wearing, what they're doing, what they're saying (if applicable), setting details, objects shown, and any visual elements.
             
-            4. Be specific about:
-               - Character descriptions (appearance, clothing, expressions)
-               - Actions and movements
-               - Objects, products, or props being displayed
-               - Setting and background details
-               - Any text, graphics, or overlays
-               - Transitions between scenes
-            
+                      
             5. Write descriptions as if you're watching the video in real-time, noting everything visible and audible.
 
-            6.Dialogue:
-            - Include short direct speech/dialogue wherever possible.
-            -  If no exact lines are known, intelligently infer short probable phrases (e.g., "Let's get started!", "Here's how you do it.", etc.)
-            
+                        
             Based on the title and description, intelligently infer what would likely happen in each time segment. Consider the video type and create contextually appropriate, detailed descriptions.
             """