File size: 11,476 Bytes
338d95d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 |
"""
Audio processing utilities for CompI Phase 2.A: Audio Input Integration
This module provides comprehensive audio analysis capabilities including:
- Audio feature extraction (tempo, energy, spectral features)
- Audio preprocessing and normalization
- Audio-to-text captioning using OpenAI Whisper
- Multimodal prompt fusion combining audio features with text prompts
"""
import os
import numpy as np
import librosa
import soundfile as sf
from typing import Dict, List, Optional, Tuple, Union
import logging
from dataclasses import dataclass
# Setup logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
@dataclass
class AudioFeatures:
"""Container for extracted audio features"""
tempo: float
energy: float # RMS energy
zero_crossing_rate: float
spectral_centroid: float
spectral_rolloff: float
mfcc_mean: np.ndarray
chroma_mean: np.ndarray
duration: float
sample_rate: int
def to_dict(self) -> Dict:
"""Convert to dictionary for JSON serialization"""
return {
'tempo': float(self.tempo),
'energy': float(self.energy),
'zero_crossing_rate': float(self.zero_crossing_rate),
'spectral_centroid': float(self.spectral_centroid),
'spectral_rolloff': float(self.spectral_rolloff),
'mfcc_mean': self.mfcc_mean.tolist() if hasattr(self.mfcc_mean, 'tolist') else list(self.mfcc_mean),
'chroma_mean': self.chroma_mean.tolist() if hasattr(self.chroma_mean, 'tolist') else list(self.chroma_mean),
'duration': float(self.duration),
'sample_rate': int(self.sample_rate)
}
class AudioProcessor:
"""Comprehensive audio processing and analysis"""
def __init__(self, target_sr: int = 16000, max_duration: float = 60.0):
"""
Initialize audio processor
Args:
target_sr: Target sample rate for processing
max_duration: Maximum audio duration to process (seconds)
"""
self.target_sr = target_sr
self.max_duration = max_duration
def load_audio(self, audio_path: str) -> Tuple[np.ndarray, int]:
"""
Load and preprocess audio file
Args:
audio_path: Path to audio file
Returns:
Tuple of (audio_data, sample_rate)
"""
try:
# Load audio with librosa
audio, sr = librosa.load(
audio_path,
sr=self.target_sr,
duration=self.max_duration
)
# Normalize audio
audio = librosa.util.normalize(audio)
logger.info(f"Loaded audio: {audio_path}, duration: {len(audio)/sr:.2f}s")
return audio, sr
except Exception as e:
logger.error(f"Error loading audio {audio_path}: {e}")
raise
def extract_features(self, audio: np.ndarray, sr: int) -> AudioFeatures:
"""
Extract comprehensive audio features
Args:
audio: Audio signal
sr: Sample rate
Returns:
AudioFeatures object containing all extracted features
"""
try:
# Basic features
duration = len(audio) / sr
# Tempo and beat tracking
tempo, _ = librosa.beat.beat_track(y=audio, sr=sr)
# Energy (RMS)
rms = librosa.feature.rms(y=audio)[0]
energy = np.sqrt(np.mean(rms**2))
# Zero crossing rate
zcr = librosa.feature.zero_crossing_rate(audio)[0]
zcr_mean = np.mean(zcr)
# Spectral features
spectral_centroids = librosa.feature.spectral_centroid(y=audio, sr=sr)[0]
spectral_centroid = np.mean(spectral_centroids)
spectral_rolloff = librosa.feature.spectral_rolloff(y=audio, sr=sr)[0]
spectral_rolloff_mean = np.mean(spectral_rolloff)
# MFCC features
mfccs = librosa.feature.mfcc(y=audio, sr=sr, n_mfcc=13)
mfcc_mean = np.mean(mfccs, axis=1)
# Chroma features
chroma = librosa.feature.chroma_stft(y=audio, sr=sr)
chroma_mean = np.mean(chroma, axis=1)
features = AudioFeatures(
tempo=float(tempo),
energy=float(energy),
zero_crossing_rate=float(zcr_mean),
spectral_centroid=float(spectral_centroid),
spectral_rolloff=float(spectral_rolloff_mean),
mfcc_mean=mfcc_mean,
chroma_mean=chroma_mean,
duration=float(duration),
sample_rate=int(sr)
)
logger.info(f"Extracted features: tempo={float(tempo):.1f}, energy={float(energy):.4f}")
return features
except Exception as e:
logger.error(f"Error extracting audio features: {e}")
raise
def analyze_audio_file(self, audio_path: str) -> AudioFeatures:
"""
Complete audio analysis pipeline
Args:
audio_path: Path to audio file
Returns:
AudioFeatures object
"""
audio, sr = self.load_audio(audio_path)
return self.extract_features(audio, sr)
class AudioCaptioner:
"""Audio-to-text captioning using OpenAI Whisper"""
def __init__(self, model_size: str = "base", device: str = "auto"):
"""
Initialize audio captioner
Args:
model_size: Whisper model size (tiny, base, small, medium, large)
device: Device to run on (auto, cpu, cuda)
"""
self.model_size = model_size
self.device = device
self._model = None
def _load_model(self):
"""Lazy load Whisper model"""
if self._model is None:
try:
import whisper
self._model = whisper.load_model(self.model_size, device=self.device)
logger.info(f"Loaded Whisper model: {self.model_size}")
except ImportError:
logger.error("OpenAI Whisper not installed. Install with: pip install openai-whisper")
raise
except Exception as e:
logger.error(f"Error loading Whisper model: {e}")
raise
def caption_audio(self, audio_path: str, language: str = "en") -> str:
"""
Generate text caption from audio
Args:
audio_path: Path to audio file
language: Language code for transcription
Returns:
Text caption of the audio content
"""
self._load_model()
try:
import whisper
# Load and preprocess audio for Whisper
audio = whisper.load_audio(audio_path)
audio = whisper.pad_or_trim(audio)
# Generate mel spectrogram
mel = whisper.log_mel_spectrogram(audio).to(self._model.device)
# Decode audio
options = whisper.DecodingOptions(language=language, fp16=False)
result = whisper.decode(self._model, mel, options)
caption = result.text.strip()
logger.info(f"Generated audio caption: '{caption[:50]}...'")
return caption
except Exception as e:
logger.error(f"Error captioning audio: {e}")
return ""
class MultimodalPromptFusion:
"""Intelligent fusion of text prompts with audio features and captions"""
def __init__(self):
"""Initialize prompt fusion system"""
pass
def fuse_prompt_with_audio(
self,
text_prompt: str,
style: str,
mood: str,
audio_features: AudioFeatures,
audio_caption: str = ""
) -> str:
"""
Create enhanced prompt by fusing text with audio analysis
Args:
text_prompt: Original text prompt
style: Art style
mood: Mood/atmosphere
audio_features: Extracted audio features
audio_caption: Audio caption from Whisper
Returns:
Enhanced multimodal prompt
"""
# Start with base prompt
enhanced_prompt = text_prompt.strip()
# Add style and mood
if style:
enhanced_prompt += f", {style}"
if mood:
enhanced_prompt += f", {mood}"
# Add audio caption if available
if audio_caption:
enhanced_prompt += f", inspired by the sound of: {audio_caption}"
# Add tempo-based descriptors
if audio_features.tempo < 80:
enhanced_prompt += ", slow and contemplative"
elif audio_features.tempo > 140:
enhanced_prompt += ", fast-paced and energetic"
elif audio_features.tempo > 120:
enhanced_prompt += ", upbeat and dynamic"
# Add energy-based descriptors
if audio_features.energy > 0.05:
enhanced_prompt += ", vibrant and powerful"
elif audio_features.energy < 0.02:
enhanced_prompt += ", gentle and subtle"
# Add rhythm-based descriptors
if audio_features.zero_crossing_rate > 0.15:
enhanced_prompt += ", rhythmic and percussive"
# Add tonal descriptors based on spectral features
if audio_features.spectral_centroid > 3000:
enhanced_prompt += ", bright and crisp"
elif audio_features.spectral_centroid < 1500:
enhanced_prompt += ", warm and deep"
logger.info(f"Enhanced prompt: {enhanced_prompt}")
return enhanced_prompt
def generate_audio_tags(self, audio_features: AudioFeatures) -> List[str]:
"""
Generate descriptive tags based on audio features
Args:
audio_features: Extracted audio features
Returns:
List of descriptive tags
"""
tags = []
# Tempo tags
if audio_features.tempo < 60:
tags.append("very_slow")
elif audio_features.tempo < 90:
tags.append("slow")
elif audio_features.tempo < 120:
tags.append("moderate")
elif audio_features.tempo < 140:
tags.append("fast")
else:
tags.append("very_fast")
# Energy tags
if audio_features.energy > 0.06:
tags.append("high_energy")
elif audio_features.energy > 0.03:
tags.append("medium_energy")
else:
tags.append("low_energy")
# Rhythm tags
if audio_features.zero_crossing_rate > 0.15:
tags.append("percussive")
elif audio_features.zero_crossing_rate < 0.05:
tags.append("smooth")
# Spectral tags
if audio_features.spectral_centroid > 3000:
tags.append("bright")
elif audio_features.spectral_centroid < 1500:
tags.append("dark")
return tags
|