funtool

app.py

@@ -1950,6 +1950,10 @@ The lyrics should:
 - Be completely original
 - Match the song duration of {duration:.1f} seconds
 
+IMPORTANT: Your generated lyrics must be followed by a section titled "[RHYTHM_ANALYSIS_SECTION]" 
+where you analyze how well the lyrics align with the musical rhythm. This section MUST appear
+even if there are no rhythm issues.
+
 Your lyrics:
 """
     else:
@@ -1996,6 +2000,10 @@ The lyrics should:
 DON'T include any section labels like [Verse] or [Chorus] unless specifically instructed.
 Instead, write lyrics that flow naturally and match the music's rhythm precisely.
 
+IMPORTANT: Your generated lyrics must be followed by a section titled "[RHYTHM_ANALYSIS_SECTION]" 
+where you analyze how well the lyrics align with the musical rhythm. This section MUST appear
+even if there are no rhythm issues.
+
 Your lyrics:
 """
 
@@ -2061,7 +2069,7 @@ Your lyrics:
         for model_name, token_id in think_end_tokens.items():
             if token_id in output_ids:
                 found_token = token_id
-                token_position = len(output_ids) - output_ids[::-1].index(token_id)
+                token_position = output_ids.index(token_id) + 1
                 break
                 
         # Use the position of the thinking token if found
@@ -2179,15 +2187,23 @@ Improved lyrics with fixed rhythm:
             
         lyrics = '\n'.join(formatted_lyrics)
     
-    # Clean up the output if there are analytical notes
-    if "[Note: Potential rhythm mismatches" in lyrics and "[How to fix rhythm mismatches" in lyrics:
-        # Optionally separate the analysis from the final lyrics for cleaner display
-        clean_lyrics = lyrics.split("[Note:")[0].strip()
-        analysis_notes = lyrics.split("[Note:")[1]
+    # Check if we have the [RHYTHM_ANALYSIS_SECTION] tag
+    if "[RHYTHM_ANALYSIS_SECTION]" in lyrics:
+        # Split at our custom marker
+        parts = lyrics.split("[RHYTHM_ANALYSIS_SECTION]")
+        clean_lyrics = parts[0].strip()
+        rhythm_analysis = parts[1].strip()
         
-        # For now, keep the full output with notes for debugging
-        # In a production system, you might want to handle this differently
-        lyrics = lyrics
+        # Add our standard marker for compatibility with existing code
+        lyrics = clean_lyrics + "\n\n[Note: Rhythm Analysis]\n" + rhythm_analysis
+    
+    # For backwards compatibility - if we have the old format, still handle it
+    elif "[Note: Potential rhythm mismatches" in lyrics:
+        # Keep it as is, the existing parsing code can handle this format
+        pass
+    else:
+        # No analysis found, add a minimal one
+        lyrics = lyrics + "\n\n[Note: Rhythm Analysis]\nNo rhythm issues detected. All syllables align well with the beat pattern."
     
     return lyrics
 
@@ -2262,7 +2278,13 @@ def process_audio(audio_file):
         }
         
         # Extract rhythm analysis if present in the lyrics
-        if isinstance(lyrics, str) and "[Note: Potential rhythm mismatches" in lyrics:
+        if isinstance(lyrics, str) and "[Note: Rhythm Analysis]" in lyrics:
+            clean_lyrics = lyrics.split("[Note: Rhythm Analysis]")[0].strip()
+            rhythm_analysis = "[Note: Rhythm Analysis]" + lyrics.split("[Note: Rhythm Analysis]")[1]
+            results["clean_lyrics"] = clean_lyrics
+            results["rhythm_analysis"] = rhythm_analysis
+        # Backwards compatibility with old format
+        elif isinstance(lyrics, str) and "[Note: Potential rhythm mismatches" in lyrics:
             clean_lyrics = lyrics.split("[Note:")[0].strip()
             rhythm_analysis = "[Note:" + lyrics.split("[Note:")[1]
             results["clean_lyrics"] = clean_lyrics
@@ -2349,9 +2371,15 @@ with gr.Blocks(title="Music Genre Classifier & Lyrics Generator") as demo:
                 
                 # Extract rhythm analysis if present
                 rhythm_analysis = "No detailed rhythm analysis available"
-                if isinstance(lyrics, str) and "[Note: Potential rhythm mismatches" in lyrics:
-                    clean_lyrics = lyrics.split("[Note:")[0].strip()
-                    rhythm_analysis = "[Note:" + lyrics.split("[Note:")[1]
+                if isinstance(lyrics, str):
+                    # First check for new format
+                    if "[Note: Rhythm Analysis]" in lyrics:
+                        clean_lyrics = lyrics.split("[Note: Rhythm Analysis]")[0].strip()
+                        rhythm_analysis = lyrics.split("[Note: Rhythm Analysis]")[1]
+                    # Check for old format
+                    elif "[Note: Potential rhythm mismatches" in lyrics:
+                        clean_lyrics = lyrics.split("[Note:")[0].strip()
+                        rhythm_analysis = "[Note:" + lyrics.split("[Note:")[1]
             
             # Format emotion analysis results
             try:
@@ -2376,17 +2404,17 @@ with gr.Blocks(title="Music Genre Classifier & Lyrics Generator") as demo:
                         else:
                             emotion_text += f"~{section['syllable_count']} syllables)\n"
                             
-                    # Add flexible structure info if available
-                    if "flexible_structure" in song_structure and song_structure["flexible_structure"]:
-                        flexible = song_structure["flexible_structure"]
-                        if "segments" in flexible and flexible["segments"]:
-                            emotion_text += "\nDetailed Rhythm Analysis:\n"
-                            for i, segment in enumerate(flexible["segments"][:5]):  # Show first 5 segments
-                                emotion_text += f"- Segment {i+1}: {segment['start']:.1f}s to {segment['end']:.1f}s, "
-                                emotion_text += f"pattern: {segment.get('syllable_template', 'N/A')}\n"
-                            
-                            if len(flexible["segments"]) > 5:
-                                emotion_text += f"  (+ {len(flexible['segments']) - 5} more segments)\n"
+                        # Add flexible structure info if available
+                        if "flexible_structure" in song_structure and song_structure["flexible_structure"]:
+                            flexible = song_structure["flexible_structure"]
+                            if "segments" in flexible and flexible["segments"]:
+                                emotion_text += "\nDetailed Rhythm Analysis:\n"
+                                for i, segment in enumerate(flexible["segments"][:5]):  # Show first 5 segments
+                                    emotion_text += f"- Segment {i+1}: {segment['start']:.1f}s to {segment['end']:.1f}s, "
+                                    emotion_text += f"pattern: {segment.get('syllable_template', 'N/A')}\n"
+                                
+                                if len(flexible["segments"]) > 5:
+                                    emotion_text += f"  (+ {len(flexible['segments']) - 5} more segments)\n"
                         
                 except Exception as e:
                     print(f"Error displaying song structure: {str(e)}")