ss

utils.py

@@ -37,40 +37,6 @@ def extract_mfcc_features(y, sr, n_mfcc=20):
         # Return a fallback feature vector if extraction fails
         return np.zeros(n_mfcc)
 
-def calculate_lyrics_length(duration):
-    """
-    Calculate appropriate lyrics length based on audio duration.
-    Uses a more conservative calculation that generates shorter lyrics:
-    - Average words per line (8-10 words)
-    - Reduced words per minute (45 words instead of 135)
-    - Simplified song structure
-    """
-    # Convert duration to minutes
-    duration_minutes = duration / 60
-    
-    # Calculate total words based on duration
-    # Using 45 words per minute (reduced from 135)
-    total_words = int(duration_minutes * 90)
-    
-    # Calculate number of lines
-    # Assuming 8-10 words per line
-    words_per_line = 9  # average
-    total_lines = total_words // words_per_line
-    
-    # Adjust for song structure with shorter lengths
-    if total_lines < 6:
-        # Very short song - keep it simple
-        return max(2, total_lines)
-    elif total_lines < 10:
-        # Short song - one verse and chorus
-        return min(6, total_lines)
-    elif total_lines < 15:
-        # Medium song - two verses and chorus
-        return min(10, total_lines)
-    else:
-        # Longer song - two verses, chorus, and bridge
-        return min(15, total_lines)
-
 def format_genre_results(top_genres):
     """Format genre classification results for display."""
     result = "Top Detected Genres:\n"
@@ -89,17 +55,3 @@ def ensure_cuda_availability():
         print("CUDA is not available. Using CPU for inference.")
     return cuda_available
 
-def preprocess_audio_for_model(waveform, sample_rate, target_sample_rate=16000, max_length=16000):
-    """Preprocess audio for model input (resample, pad/trim)."""
-    # Resample if needed
-    if sample_rate != target_sample_rate:
-        waveform = librosa.resample(waveform, orig_sr=sample_rate, target_sr=target_sample_rate)
-    
-    # Trim or pad to expected length
-    if len(waveform) > max_length:
-        waveform = waveform[:max_length]
-    elif len(waveform) < max_length:
-        padding = max_length - len(waveform)
-        waveform = np.pad(waveform, (0, padding), 'constant')
-        
-    return waveform