Update app.py

app.py

@@ -172,27 +172,27 @@ def get_image_prompt(sentiment, transcribed_text, chunk_idx, total_chunks):
     base_prompt = f"Chunk {chunk_idx+1}/{total_chunks}: "
     
     if sentiment == "positive":
-        return   f"Generate a black and white non-figurative equirectangular 360 abstract sketch drawing of:{transcribed_text}. Add light glow colors over the b/w sketch with low histogram frequency in bright bins, dominant color in high RGB range, and high brightness and color variance. Apply high-frequency texture with strong filter energy, pronounced gradient magnitude, and strong local contrast. Use high spatial complexity, increased horizontal and vertical symmetry, high edge density, bright gray levels, and high contrast. Emphasize rich visual structure, color variation, and texture intensity across spatial composition."
+        return   f"Generate a black and white equirectangular sketch drawing with light glow colors details of:{transcribed_text}. Use low histogram frequency in bright bins, dominant color in high RGB range, and high brightness and color variance. Apply high-frequency texture with strong filter energy, pronounced gradient magnitude, and strong local contrast. Use high spatial complexity, increased horizontal and vertical symmetry, high edge density, bright gray levels, and high contrast. Emphasize rich visual structure, color variation, and texture intensity across spatial composition."
     
     elif sentiment == "negative":
-        return  f"Generate a black and white non-figurative equirectangular 360 abstract sketch drawing of:{transcribed_text}. Add light glow colors over the b/w sketch with high histogram frequency in dark bins, dominant color in low RGB range, and low brightness and color variance. Apply low-frequency texture with low filter energy, weak gradient magnitude, and low local contrast. Use low spatial complexity, reduced horizontal and vertical symmetry, low edge density, dark gray levels, and moderate contrast. Emphasize coarse structure and limited variation in color, texture, and spatial distribution."
+        return  f"Generate a black and white equirectangular sketch drawing with light glow colors details of:{transcribed_text}. Use high histogram frequency in dark bins, dominant color in low RGB range, and low brightness and color variance. Apply low-frequency texture with low filter energy, weak gradient magnitude, and low local contrast. Use low spatial complexity, reduced horizontal and vertical symmetry, low edge density, dark gray levels, and moderate contrast. Emphasize coarse structure and limited variation in color, texture, and spatial distribution."
     
     else:  # neutral
-        return  f"Generate a black and white non-figurative abstract equirectangular 360 abstract sketch drawing of:{transcribed_text}.  Add light glow colors over the b/w sketch with a balanced histogram frequency across bins, dominant color in a mid RGB range, and moderate brightness and color variance. Apply medium-frequency texture with moderate filter energy, standard gradient magnitude, and average local contrast. Use medium spatial complexity, balanced horizontal and vertical symmetry, medium edge density, mid-range gray levels, and standard contrast. Emphasize naturalistic structure and typical variation in color, texture, and spatial distribution."
+        return  f"Generate a black and white equirectangular sketch drawing with light glow colors details of:{transcribed_text}. Use a balanced histogram frequency across bins, dominant color in a mid RGB range, and moderate brightness and color variance. Apply medium-frequency texture with moderate filter energy, standard gradient magnitude, and average local contrast. Use medium spatial complexity, balanced horizontal and vertical symmetry, medium edge density, mid-range gray levels, and standard contrast. Emphasize naturalistic structure and typical variation in color, texture, and spatial distribution."
 
 # Function to get music prompt based on emotion
 def get_music_prompt(emotion, transcribed_text, chunk_idx, total_chunks):
     base_prompt = f"Chunk {chunk_idx+1}/{total_chunks}: "
     
     emotion_prompts = {
-        'neutral': f"Generate a neutral soundtrack with balanced energy and smooth spectral profile. Use moderate tempo (~100 BPM), onset rate around 2.8/sec, spectral centroid near 1000 Hz, and low dissonance. Keep pitch salience moderate (0.50) and loudness stable (~0.70 dB). Maintain low harmonic change rate (~0.05/sec) and tonal entropy 1.5 for equilibrium. Emphasize tonal balance, steady dynamics, and calm tonal centers. The music should feel even, ambient, and unobtrusive, complementing: {transcribed_text}.",
-        'calm': f"Generate a calm soundtrack with a slow tempo (~85 BPM), low onset rate (~2.2/sec), soft spectral centroid (~850 Hz), and smooth timbral evolution. Use low dissonance, high spectral flatness, and gentle pitch salience (~0.48). Keep loudness low (~0.65 dB) with infrequent harmonic changes (~0.04/sec) and stable tonality (Krumhansl value 0.80, major mode). The music should evoke tranquility and serenity through warm timbres, sustained harmonies, and flowing textures inspired by: {transcribed_text}.",      
-        'happy': f"Generate a happy soundtrack with fast tempo (~127 BPM), dense rhythmic activity (~4.2 onsets/sec), and bright timbre (spectral centroid ~1321 Hz). Use variable dissonance and peaked spectral kurtosis to create vivid texture. Maintain pitch salience (~0.54), loudness (~0.90 dB), and chord change rate (~0.07/sec). Keep tonal entropy moderate (1.95) and Krumhansl value (0.83, major mode). The music should convey joy and positivity through energetic rhythms, ornamented melodic contours, and harmonically grounded progressions inspired by: {transcribed_text}.",
-        'sad': f"Generate a sad soundtrack with slow tempo (~72 BPM), sparse onset rate (~2.0/sec), and dark timbre (spectral centroid ~720 Hz). Use moderate dissonance, low spectral kurtosis, and soft pitch salience (~0.45). Keep loudness subdued (~0.60 dB) with rare harmonic changes (~0.05/sec) and low tonal entropy (~1.4). Emphasize minor mode with gentle phrasing and sustained harmonic textures. The music should evoke sadness, intimacy, and reflection in relation to: {transcribed_text}.",
-        'angry': f"Generate an angry soundtrack with moderately fast tempo (~120 BPM), onset rate (~3.4/sec), and bright, sharp timbre (spectral centroid ~2002 Hz). Use flat spectral kurtosis and stable dissonance. Maintain clear pitch salience (~0.58), high loudness (~0.96 dB), and frequent chord changes (~0.10/sec). Set tonal entropy to 2.57 and Krumhansl key profile (~0.54, minor mode). The music should express anger through strong rhythmic drive, aggressive articulation, and harmonically unstable progressions that reflect: {transcribed_text}.",
-        'fearful': f"Generate a fearful soundtrack with irregular tempo (~95 BPM), fluctuating onset rate (~3.0/sec), and high spectral variability (centroid ~1750 Hz). Use unstable dissonance, low pitch salience (~0.42), and dynamic loudness (~0.80 dB). Increase chord change irregularity (~0.09/sec) and tonal entropy (2.4, minor mode). Emphasize eerie textures, spatial tension, and spectral modulation. The music should evoke suspense, fear, and anticipation inspired by: {transcribed_text}.",
-        'disgust': f"Generate a disgusted soundtrack with moderate tempo (~90 BPM), irregular onset rate (~2.5/sec), and dark, rough timbre (spectral centroid ~950 Hz). Use dissonant harmonies, unstable spectral kurtosis, and low pitch salience (~0.40). Keep loudness (~0.75 dB) and tonal entropy (~2.2, minor mode). The music should evoke discomfort and unease through distorted textures, rough intervals, and unstable harmonic motion reflecting: {transcribed_text}.",
-        'surprised': f"Generate a surprised soundtrack with variable tempo (~110 BPM), fluctuating onset rate (~3.8/sec), and dynamic spectral centroid (~1500 Hz). Use high spectral kurtosis and pitch salience (~0.57) to accent sudden contrasts. Loudness should vary (~0.85 dB) with irregular chord changes (~0.11/sec) and moderate tonal entropy (~2.0, major mode). The music should evoke surprise and wonder through abrupt transitions, playful motifs, and expressive timbral changes inspired by: {transcribed_text}."
+        'neutral': f"Generate a neutral orchestral soundtrack with balanced energy and smooth spectral profile. Use moderate tempo (~100 BPM), onset rate around 2.8/sec, spectral centroid near 1000 Hz, and low dissonance. Keep pitch salience moderate (0.50) and loudness stable (~0.70 dB). Maintain low harmonic change rate (~0.05/sec) and tonal entropy 1.5 for equilibrium. Emphasize tonal balance, steady dynamics, and calm tonal centers. The music should feel even, ambient, and unobtrusive, complementing: {transcribed_text}.",
+        'calm': f"Generate a calm orchestral soundtrack with a slow tempo (~85 BPM), low onset rate (~2.2/sec), soft spectral centroid (~850 Hz), and smooth timbral evolution. Use low dissonance, high spectral flatness, and gentle pitch salience (~0.48). Keep loudness low (~0.65 dB) with infrequent harmonic changes (~0.04/sec) and stable tonality (Krumhansl value 0.80, major mode). The music should evoke tranquility and serenity through warm timbres, sustained harmonies, and flowing textures inspired by: {transcribed_text}.",      
+        'happy': f"Generate a happy orchestral soundtrack with fast tempo (~127 BPM), dense rhythmic activity (~4.2 onsets/sec), and bright timbre (spectral centroid ~1321 Hz). Use variable dissonance and peaked spectral kurtosis to create vivid texture. Maintain pitch salience (~0.54), loudness (~0.90 dB), and chord change rate (~0.07/sec). Keep tonal entropy moderate (1.95) and Krumhansl value (0.83, major mode). The music should convey joy and positivity through energetic rhythms, ornamented melodic contours, and harmonically grounded progressions inspired by: {transcribed_text}.",
+        'sad': f"Generate a sad orchestral soundtrack with slow tempo (~72 BPM), sparse onset rate (~2.0/sec), and dark timbre (spectral centroid ~720 Hz). Use moderate dissonance, low spectral kurtosis, and soft pitch salience (~0.45). Keep loudness subdued (~0.60 dB) with rare harmonic changes (~0.05/sec) and low tonal entropy (~1.4). Emphasize minor mode with gentle phrasing and sustained harmonic textures. The music should evoke sadness, intimacy, and reflection in relation to: {transcribed_text}.",
+        'angry': f"Generate an angry orchestral soundtrack with moderately fast tempo (~120 BPM), onset rate (~3.4/sec), and bright, sharp timbre (spectral centroid ~2002 Hz). Use flat spectral kurtosis and stable dissonance. Maintain clear pitch salience (~0.58), high loudness (~0.96 dB), and frequent chord changes (~0.10/sec). Set tonal entropy to 2.57 and Krumhansl key profile (~0.54, minor mode). The music should express anger through strong rhythmic drive, aggressive articulation, and harmonically unstable progressions that reflect: {transcribed_text}.",
+        'fearful': f"Generate a orchestral fearful soundtrack with irregular tempo (~95 BPM), fluctuating onset rate (~3.0/sec), and high spectral variability (centroid ~1750 Hz). Use unstable dissonance, low pitch salience (~0.42), and dynamic loudness (~0.80 dB). Increase chord change irregularity (~0.09/sec) and tonal entropy (2.4, minor mode). Emphasize eerie textures, spatial tension, and spectral modulation. The music should evoke suspense, fear, and anticipation inspired by: {transcribed_text}.",
+        'disgust': f"Generate a orchestral disgusted soundtrack with moderate tempo (~90 BPM), irregular onset rate (~2.5/sec), and dark, rough timbre (spectral centroid ~950 Hz). Use dissonant harmonies, unstable spectral kurtosis, and low pitch salience (~0.40). Keep loudness (~0.75 dB) and tonal entropy (~2.2, minor mode). The music should evoke discomfort and unease through distorted textures, rough intervals, and unstable harmonic motion reflecting: {transcribed_text}.",
+        'surprised': f"Generate a orchestral surprised soundtrack with variable tempo (~110 BPM), fluctuating onset rate (~3.8/sec), and dynamic spectral centroid (~1500 Hz). Use high spectral kurtosis and pitch salience (~0.57) to accent sudden contrasts. Loudness should vary (~0.85 dB) with irregular chord changes (~0.11/sec) and moderate tonal entropy (~2.0, major mode). The music should evoke surprise and wonder through abrupt transitions, playful motifs, and expressive timbral changes inspired by: {transcribed_text}."
     }
     
     return emotion_prompts.get(