Update app.py
Browse files
app.py
CHANGED
@@ -1,205 +1,454 @@
|
|
1 |
import gradio as gr
|
2 |
-
|
3 |
-
import
|
4 |
-
import
|
5 |
import librosa
|
|
|
|
|
6 |
import numpy as np
|
7 |
-
|
8 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
import shutil
|
10 |
-
import
|
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 |
-
audio_file = output_dir / f"{info['title']}.mp3"
|
68 |
-
logging.debug(f"Downloaded audio: {audio_file}")
|
69 |
-
output_audio = str(audio_file)
|
70 |
-
|
71 |
-
# Perform automatic audio feature analysis with librosa
|
72 |
-
y, sr = librosa.load(audio_file)
|
73 |
-
hop_length = 512 # Valid hop_length to fix "Invalid hop_length: 0" error
|
74 |
-
logging.debug(f"Using hop_length: {hop_length}")
|
75 |
-
|
76 |
-
# Extract features
|
77 |
-
mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13, hop_length=hop_length)
|
78 |
-
spectral_centroid = librosa.feature.spectral_centroid(y=y, sr=sr, hop_length=hop_length)
|
79 |
-
tempo, _ = librosa.beat.beat_track(y=y, sr=sr, hop_length=hop_length)
|
80 |
-
|
81 |
-
# Aggregate features
|
82 |
-
mfcc_mean = np.mean(mfcc, axis=1).tolist()[:3] # Mean of first 3 MFCC coefficients
|
83 |
-
spectral_centroid_mean = np.mean(spectral_centroid)
|
84 |
-
features_summary = (
|
85 |
-
f"Audio Features: MFCC (mean of first 3 coeffs): {mfcc_mean}, "
|
86 |
-
f"Spectral Centroid: {spectral_centroid_mean:.2f} Hz, "
|
87 |
-
f"Tempo: {tempo:.2f} BPM"
|
88 |
-
)
|
89 |
-
|
90 |
-
processed_text += f" {features_summary}."
|
91 |
-
extra_info += f", Audio: {audio_file.name}"
|
92 |
-
|
93 |
-
except Exception as e:
|
94 |
-
logging.error(f"YouTube download or audio processing error: {str(e)}")
|
95 |
-
processed_text += f" Error processing YouTube audio: {str(e)}."
|
96 |
-
|
97 |
-
# Handle image processing if provided
|
98 |
-
if input_image is not None:
|
99 |
-
from PIL import ImageEnhance
|
100 |
-
enhancer = ImageEnhance.Brightness(input_image)
|
101 |
-
output_image_display = enhancer.enhance(1.5)
|
102 |
-
processed_text += " Image processed (brightened)."
|
103 |
-
else:
|
104 |
-
processed_text += " No image provided."
|
105 |
-
|
106 |
-
# Incorporate slider and checkbox
|
107 |
-
processed_text += f" Slider: {slider_value}, Enhanced Analysis: {checkbox_value}."
|
108 |
-
if checkbox_value:
|
109 |
-
processed_text += " Enhanced analysis enabled."
|
110 |
-
if youtube_url and slider_value > 50:
|
111 |
-
processed_text += f" High threshold ({slider_value}) applied for deeper analysis."
|
112 |
-
|
113 |
-
return processed_text, output_image_display, output_audio, extra_info
|
114 |
-
|
115 |
-
except Exception as e:
|
116 |
-
logging.error(f"Error in analyze_audio: {str(e)}")
|
117 |
-
return f"Error: {str(e)}", None, None, "Processing failed."
|
118 |
-
|
119 |
-
finally:
|
120 |
-
# Clean up temporary directory after a delay to ensure file access
|
121 |
try:
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
|
|
|
|
|
|
|
|
126 |
except Exception as e:
|
127 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
128 |
|
129 |
-
# Define input components
|
130 |
-
input_youtube_url = gr.Textbox(
|
131 |
-
label="YouTube Video URL",
|
132 |
-
placeholder="e.g., https://www.youtube.com/watch?v=dQw4w9WgXcQ",
|
133 |
-
info="Optional: Enter a YouTube URL to download and analyze audio."
|
134 |
-
)
|
135 |
-
input_text_component = gr.Textbox(
|
136 |
-
label="Input Text",
|
137 |
-
placeholder="e.g., Analyze this audio track",
|
138 |
-
info="Type a description or query for processing."
|
139 |
-
)
|
140 |
-
input_image_component = gr.Image(
|
141 |
-
type="pil",
|
142 |
-
label="Upload Image (Optional)",
|
143 |
-
sources=["upload", "webcam", "clipboard"]
|
144 |
-
)
|
145 |
-
input_slider_component = gr.Slider(
|
146 |
-
minimum=0,
|
147 |
-
maximum=100,
|
148 |
-
value=50,
|
149 |
-
step=1,
|
150 |
-
label="Analysis Threshold",
|
151 |
-
info="Adjusts sensitivity of audio feature analysis."
|
152 |
-
)
|
153 |
-
input_checkbox_component = gr.Checkbox(
|
154 |
-
label="Enable Enhanced Analysis",
|
155 |
-
info="Toggle for deeper audio feature extraction."
|
156 |
-
)
|
157 |
|
158 |
-
# Define output components
|
159 |
-
output_text_component = gr.Textbox(
|
160 |
-
label="Analysis Results",
|
161 |
-
info="Text results including audio feature analysis."
|
162 |
-
)
|
163 |
-
output_image_component = gr.Image(
|
164 |
-
label="Processed Image (if any)",
|
165 |
-
info="Processed image output (if provided)."
|
166 |
-
)
|
167 |
-
output_audio_component = gr.Audio(
|
168 |
-
label="Downloaded Audio",
|
169 |
-
type="filepath",
|
170 |
-
info="Audio downloaded from YouTube."
|
171 |
-
)
|
172 |
-
output_label_component = gr.Label(
|
173 |
-
label="Analysis Summary",
|
174 |
-
info="Feature analysis details and processing info."
|
175 |
-
)
|
176 |
|
177 |
-
# Create the Gradio interface
|
178 |
-
iface = gr.Interface(
|
179 |
-
fn=analyze_audio,
|
180 |
-
inputs=[
|
181 |
-
input_youtube_url,
|
182 |
-
input_text_component,
|
183 |
-
input_image_component,
|
184 |
-
input_slider_component,
|
185 |
-
input_checkbox_component
|
186 |
-
],
|
187 |
-
outputs=[
|
188 |
-
output_text_component,
|
189 |
-
output_image_component,
|
190 |
-
output_audio_component,
|
191 |
-
output_label_component
|
192 |
-
],
|
193 |
-
title="YouTube Audio Feature Analysis",
|
194 |
-
description="Download YouTube audio, analyze features with librosa, and process text/image inputs. Customize with slider and checkbox.",
|
195 |
-
examples=[
|
196 |
-
["https://www.youtube.com/watch?v=dQw4w9WgXcQ", "Analyze this track", None, 75, True],
|
197 |
-
[None, "Describe a music track", None, 30, False],
|
198 |
-
["https://www.youtube.com/watch?v=9bZkp7q19f0", "Extract audio features", None, 60, True]
|
199 |
-
],
|
200 |
-
allow_flagging="never",
|
201 |
-
theme=gr.themes.Soft()
|
202 |
-
)
|
203 |
|
204 |
if __name__ == "__main__":
|
205 |
-
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import subprocess
|
3 |
+
import os
|
4 |
+
import tempfile
|
5 |
import librosa
|
6 |
+
import librosa.display
|
7 |
+
import matplotlib.pyplot as plt
|
8 |
import numpy as np
|
9 |
+
import scipy.ndimage
|
10 |
+
from pathlib import Path
|
11 |
+
|
12 |
+
import logging
|
13 |
+
import warnings
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
import shutil
|
18 |
+
from typing import Tuple, Optional, Dict, Any
|
19 |
+
|
20 |
+
# Configure matplotlib for web display
|
21 |
+
plt.switch_backend('Agg')
|
22 |
+
warnings.filterwarnings('ignore')
|
23 |
+
|
24 |
+
# Setup logging
|
25 |
+
logging.basicConfig(
|
26 |
+
level=logging.INFO,
|
27 |
+
format="%(asctime)s - %(levelname)s - %(message)s",
|
28 |
+
handlers=[logging.StreamHandler()]
|
29 |
+
)
|
30 |
+
logger = logging.getLogger(__name__)
|
31 |
+
|
32 |
+
class AudioAnalyzer:
|
33 |
+
"""Core class for audio analysis with modular feature extraction methods."""
|
34 |
+
|
35 |
+
def __init__(self, temp_dir: Optional[str] = None):
|
36 |
+
"""Initialize with a temporary directory for file storage."""
|
37 |
+
self.temp_dir = Path(temp_dir or tempfile.mkdtemp())
|
38 |
+
self.temp_dir.mkdir(exist_ok=True)
|
39 |
+
self.plot_files = [] # Track plot files for cleanup
|
40 |
+
logger.info(f"Initialized temporary directory: {self.temp_dir}")
|
41 |
+
|
42 |
+
def cleanup(self) -> None:
|
43 |
+
"""Remove temporary directory and plot files."""
|
44 |
+
for plot_file in self.plot_files:
|
45 |
+
if Path(plot_file).exists():
|
46 |
+
try:
|
47 |
+
Path(plot_file).unlink()
|
48 |
+
logger.info(f"Removed plot file: {plot_file}")
|
49 |
+
except Exception as e:
|
50 |
+
logger.warning(f"Failed to remove plot file {plot_file}: {str(e)}")
|
51 |
+
if self.temp_dir.exists():
|
52 |
+
shutil.rmtree(self.temp_dir, ignore_errors=True)
|
53 |
+
logger.info(f"Cleaned up temporary directory: {self.temp_dir}")
|
54 |
+
|
55 |
+
def download_youtube_audio(self, video_url: str, progress=gr.Progress()) -> Tuple[Optional[str], str]:
|
56 |
+
"""Download audio from YouTube using yt-dlp."""
|
57 |
+
if not video_url:
|
58 |
+
return None, "Please provide a valid YouTube URL"
|
59 |
+
|
60 |
+
progress(0.1, desc="Initializing download...")
|
61 |
+
output_dir = self.temp_dir / "downloaded_audio"
|
62 |
+
output_dir.mkdir(exist_ok=True)
|
63 |
+
output_file = output_dir / "audio.mp3"
|
64 |
+
|
65 |
+
command = [
|
66 |
+
"yt-dlp",
|
67 |
+
"-x",
|
68 |
+
"--audio-format", "mp3",
|
69 |
+
"-o", str(output_file),
|
70 |
+
"--no-playlist",
|
71 |
+
"--restrict-filenames",
|
72 |
+
video_url
|
73 |
+
]
|
74 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
try:
|
76 |
+
progress(0.3, desc="Downloading audio...")
|
77 |
+
subprocess.run(command, check=True, capture_output=True, text=True)
|
78 |
+
progress(1.0, desc="Download complete!")
|
79 |
+
return str(output_file), f"Successfully downloaded audio: {output_file.name}"
|
80 |
+
except FileNotFoundError:
|
81 |
+
return None, "yt-dlp not found. Install it with: pip install yt-dlp"
|
82 |
+
except subprocess.CalledProcessError as e:
|
83 |
+
return None, f"Download failed: {e.stderr}"
|
84 |
except Exception as e:
|
85 |
+
logger.error(f"Unexpected error during download: {str(e)}")
|
86 |
+
return None, f"Error: {str(e)}"
|
87 |
+
|
88 |
+
def save_plot(self, fig, filename: str) -> Optional[str]:
|
89 |
+
"""Save matplotlib figure to a temporary file and verify existence."""
|
90 |
+
try:
|
91 |
+
# Use NamedTemporaryFile to ensure persistence
|
92 |
+
with tempfile.NamedTemporaryFile(suffix='.png', delete=False, dir=self.temp_dir) as tmp_file:
|
93 |
+
plot_path = tmp_file.name
|
94 |
+
fig.savefig(plot_path, dpi=300, bbox_inches='tight', format='png')
|
95 |
+
plt.close(fig)
|
96 |
+
if not Path(plot_path).exists():
|
97 |
+
logger.error(f"Plot file not created: {plot_path}")
|
98 |
+
return None
|
99 |
+
self.plot_files.append(plot_path)
|
100 |
+
logger.info(f"Saved plot: {plot_path}")
|
101 |
+
return str(plot_path)
|
102 |
+
except Exception as e:
|
103 |
+
logger.error(f"Error saving plot {filename}: {str(e)}")
|
104 |
+
plt.close(fig)
|
105 |
+
return None
|
106 |
+
|
107 |
+
def extract_basic_features(self, audio_path: str, sr: int = 16000, max_duration: float = 60.0,
|
108 |
+
progress=gr.Progress()) -> Tuple[Optional[str], Optional[str], Optional[str]]:
|
109 |
+
"""Extract basic audio features and generate visualizations."""
|
110 |
+
if not audio_path or not Path(audio_path).exists():
|
111 |
+
return None, None, "Invalid or missing audio file"
|
112 |
+
|
113 |
+
try:
|
114 |
+
progress(0.1, desc="Loading audio...")
|
115 |
+
y, sr = librosa.load(audio_path, sr=sr)
|
116 |
+
duration = librosa.get_duration(y=y, sr=sr)
|
117 |
+
|
118 |
+
if duration > max_duration:
|
119 |
+
y = y[:int(sr * max_duration)]
|
120 |
+
duration = max_duration
|
121 |
+
|
122 |
+
progress(0.3, desc="Computing features...")
|
123 |
+
features: Dict[str, Any] = {
|
124 |
+
'duration': duration,
|
125 |
+
'sample_rate': sr,
|
126 |
+
'samples': len(y),
|
127 |
+
'tempo': float(librosa.beat.beat_track(y=y, sr=sr)[0]), # Convert to float
|
128 |
+
'mfcc': librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13),
|
129 |
+
'spectral_centroid': librosa.feature.spectral_centroid(y=y, sr=sr)[0],
|
130 |
+
'spectral_rolloff': librosa.feature.spectral_rolloff(y=y, sr=sr)[0],
|
131 |
+
'zero_crossing_rate': librosa.feature.zero_crossing_rate(y)[0]
|
132 |
+
}
|
133 |
+
|
134 |
+
progress(0.5, desc="Computing mel spectrogram...")
|
135 |
+
hop_length = 512
|
136 |
+
S_mel = librosa.feature.melspectrogram(y=y, sr=sr, hop_length=hop_length, n_mels=80)
|
137 |
+
S_dB = librosa.power_to_db(S_mel, ref=np.max)
|
138 |
+
|
139 |
+
progress(0.8, desc="Creating visualizations...")
|
140 |
+
fig, axes = plt.subplots(2, 2, figsize=(15, 10))
|
141 |
+
|
142 |
+
time_axis = np.linspace(0, duration, len(y))
|
143 |
+
axes[0, 0].plot(time_axis, y)
|
144 |
+
axes[0, 0].set_title('Waveform')
|
145 |
+
axes[0, 0].set_xlabel('Time (s)')
|
146 |
+
axes[0, 0].set_ylabel('Amplitude')
|
147 |
+
|
148 |
+
librosa.display.specshow(S_dB, sr=sr, hop_length=hop_length,
|
149 |
+
x_axis='time', y_axis='mel', ax=axes[0, 1])
|
150 |
+
axes[0, 1].set_title('Mel Spectrogram')
|
151 |
+
|
152 |
+
librosa.display.specshow(features['mfcc'], sr=sr, x_axis='time', ax=axes[1, 0])
|
153 |
+
axes[1, 0].set_title('MFCC')
|
154 |
+
|
155 |
+
times = librosa.frames_to_time(range(len(features['spectral_centroid'])), sr=sr, hop_length=hop_length)
|
156 |
+
axes[1, 1].plot(times, features['spectral_centroid'], label='Spectral Centroid')
|
157 |
+
axes[1, 1].plot(times, features['spectral_rolloff'], label='Spectral Rolloff')
|
158 |
+
axes[1, 1].set_title('Spectral Features')
|
159 |
+
axes[1, 1].set_xlabel('Time (s)')
|
160 |
+
axes[1, 1].legend()
|
161 |
+
|
162 |
+
plt.tight_layout()
|
163 |
+
plot_path = self.save_plot(fig, "basic_features")
|
164 |
+
if not plot_path:
|
165 |
+
return None, None, "Failed to save feature visualizations"
|
166 |
+
|
167 |
+
# Validate feature shapes
|
168 |
+
for key in ['mfcc', 'spectral_centroid', 'spectral_rolloff', 'zero_crossing_rate']:
|
169 |
+
if not isinstance(features[key].shape, tuple):
|
170 |
+
logger.error(f"Invalid shape for {key}: {features[key].shape}")
|
171 |
+
return None, None, f"Invalid feature shape for {key}"
|
172 |
+
|
173 |
+
summary = f"""
|
174 |
+
**Audio Summary:**
|
175 |
+
- Duration: {duration:.2f} seconds
|
176 |
+
- Sample Rate: {sr} Hz
|
177 |
+
- Estimated Tempo: {features['tempo']:.1f} BPM
|
178 |
+
- Number of Samples: {features['samples']:,}
|
179 |
+
|
180 |
+
**Feature Shapes:**
|
181 |
+
- MFCC: {features['mfcc'].shape}
|
182 |
+
- Spectral Centroid: {features['spectral_centroid'].shape}
|
183 |
+
- Spectral Rolloff: {features['spectral_rolloff'].shape}
|
184 |
+
- Zero Crossing Rate: {features['zero_crossing_rate'].shape}
|
185 |
+
"""
|
186 |
+
|
187 |
+
progress(1.0, desc="Analysis complete!")
|
188 |
+
return plot_path, summary, None
|
189 |
+
|
190 |
+
except Exception as e:
|
191 |
+
logger.error(f"Error processing audio: {str(e)}")
|
192 |
+
return None, None, f"Error processing audio: {str(e)}"
|
193 |
+
|
194 |
+
def extract_chroma_features(self, audio_path: str, sr: int = 16000, max_duration: float = 30.0,
|
195 |
+
progress=gr.Progress()) -> Tuple[Optional[str], Optional[str], Optional[str]]:
|
196 |
+
"""Extract and visualize enhanced chroma features."""
|
197 |
+
if not audio_path or not Path(audio_path).exists():
|
198 |
+
return None, None, "Invalid or missing audio file"
|
199 |
+
|
200 |
+
try:
|
201 |
+
progress(0.1, desc="Loading audio...")
|
202 |
+
y, sr = librosa.load(audio_path, sr=sr)
|
203 |
+
if len(y) > sr * max_duration:
|
204 |
+
y = y[:int(sr * max_duration)]
|
205 |
+
|
206 |
+
progress(0.3, desc="Computing chroma variants...")
|
207 |
+
chroma_orig = librosa.feature.chroma_cqt(y=y, sr=sr)
|
208 |
+
y_harm = librosa.effects.harmonic(y=y, margin=8)
|
209 |
+
chroma_harm = librosa.feature.chroma_cqt(y=y_harm, sr=sr)
|
210 |
+
chroma_filter = np.minimum(chroma_harm,
|
211 |
+
librosa.decompose.nn_filter(chroma_harm,
|
212 |
+
aggregate=np.median,
|
213 |
+
metric='cosine'))
|
214 |
+
chroma_smooth = scipy.ndimage.median_filter(chroma_filter, size=(1, 9))
|
215 |
+
chroma_stft = librosa.feature.chroma_stft(y=y, sr=sr)
|
216 |
+
chroma_cens = librosa.feature.chroma_cens(y=y, sr=sr)
|
217 |
+
|
218 |
+
progress(0.8, desc="Creating visualizations...")
|
219 |
+
fig, axes = plt.subplots(3, 2, figsize=(15, 12))
|
220 |
+
axes = axes.flatten()
|
221 |
+
|
222 |
+
for i, (chroma, title) in enumerate([
|
223 |
+
(chroma_orig, 'Original Chroma (CQT)'),
|
224 |
+
(chroma_harm, 'Harmonic Chroma'),
|
225 |
+
(chroma_filter, 'Non-local Filtered'),
|
226 |
+
(chroma_smooth, 'Median Filtered'),
|
227 |
+
(chroma_stft, 'Chroma (STFT)'),
|
228 |
+
(chroma_cens, 'CENS Features')
|
229 |
+
]):
|
230 |
+
librosa.display.specshow(chroma, y_axis='chroma', x_axis='time', ax=axes[i])
|
231 |
+
axes[i].set_title(title)
|
232 |
+
|
233 |
+
plt.tight_layout()
|
234 |
+
plot_path = self.save_plot(fig, "chroma_features")
|
235 |
+
if not plot_path:
|
236 |
+
return None, None, "Failed to save chroma visualizations"
|
237 |
+
|
238 |
+
summary = "Chroma feature analysis complete! Visualizations show different chroma extraction methods for harmonic analysis."
|
239 |
+
progress(1.0, desc="Chroma analysis complete!")
|
240 |
+
return plot_path, summary, None
|
241 |
+
|
242 |
+
except Exception as e:
|
243 |
+
logger.error(f"Error processing chroma features: {str(e)}")
|
244 |
+
return None, None, f"Error processing chroma features: {str(e)}"
|
245 |
+
|
246 |
+
def generate_patches(self, audio_path: str, sr: int = 16000, patch_duration: float = 5.0,
|
247 |
+
hop_duration: float = 1.0, progress=gr.Progress()) -> Tuple[Optional[str], Optional[str], Optional[str]]:
|
248 |
+
"""Generate fixed-duration patches for transformer input."""
|
249 |
+
if not audio_path or not Path(audio_path).exists():
|
250 |
+
return None, None, "Invalid or missing audio file"
|
251 |
+
|
252 |
+
try:
|
253 |
+
progress(0.1, desc="Loading audio...")
|
254 |
+
y, sr = librosa.load(audio_path, sr=sr)
|
255 |
+
|
256 |
+
progress(0.3, desc="Computing mel spectrogram...")
|
257 |
+
hop_length = 512
|
258 |
+
S_mel = librosa.feature.melspectrogram(y=y, sr=sr, hop_length=hop_length, n_mels=80)
|
259 |
+
S_dB = librosa.power_to_db(S_mel, ref=np.max)
|
260 |
+
|
261 |
+
progress(0.5, desc="Generating patches...")
|
262 |
+
patch_frames = librosa.time_to_frames(patch_duration, sr=sr, hop_length=hop_length)
|
263 |
+
hop_frames = librosa.time_to_frames(hop_duration, sr=sr, hop_length=hop_length)
|
264 |
+
patches = librosa.util.frame(S_dB, frame_length=patch_frames, hop_length=hop_frames)
|
265 |
+
|
266 |
+
progress(0.8, desc="Creating visualizations...")
|
267 |
+
num_patches_to_show = min(6, patches.shape[-1])
|
268 |
+
fig, axes = plt.subplots(2, 3, figsize=(18, 8))
|
269 |
+
axes = axes.flatten()
|
270 |
+
|
271 |
+
for i in range(num_patches_to_show):
|
272 |
+
librosa.display.specshow(patches[..., i], y_axis='mel', x_axis='time',
|
273 |
+
ax=axes[i], sr=sr, hop_length=hop_length)
|
274 |
+
axes[i].set_title(f'Patch {i+1}')
|
275 |
+
|
276 |
+
for i in range(num_patches_to_show, len(axes)):
|
277 |
+
axes[i].set_visible(False)
|
278 |
+
|
279 |
+
plt.tight_layout()
|
280 |
+
plot_path = self.save_plot(fig, "patches")
|
281 |
+
if not plot_path:
|
282 |
+
return None, None, "Failed to save patch visualizations"
|
283 |
+
|
284 |
+
summary = f"""
|
285 |
+
**Patch Generation Summary:**
|
286 |
+
- Total patches generated: {patches.shape[-1]}
|
287 |
+
- Patch duration: {patch_duration:.1f} seconds
|
288 |
+
- Hop duration: {hop_duration:.1f} seconds
|
289 |
+
- Patch shape (mels, time, patches): {patches.shape}
|
290 |
+
- Each patch covers {patch_frames} time frames
|
291 |
+
"""
|
292 |
+
|
293 |
+
progress(1.0, desc="Patch generation complete!")
|
294 |
+
return plot_path, summary, None
|
295 |
+
|
296 |
+
except Exception as e:
|
297 |
+
logger.error(f"Error generating patches: {str(e)}")
|
298 |
+
return None, None, f"Error generating patches: {str(e)}"
|
299 |
+
|
300 |
+
def create_gradio_interface() -> gr.Blocks:
|
301 |
+
"""Create a modular Gradio interface for audio analysis."""
|
302 |
+
analyzer = AudioAnalyzer()
|
303 |
+
|
304 |
+
with gr.Blocks(title="🎵 Audio Analysis Suite", theme=gr.themes.Soft()) as demo:
|
305 |
+
gr.Markdown("""
|
306 |
+
# 🎵 Audio Analysis Suite
|
307 |
+
|
308 |
+
Analyze audio from YouTube videos or uploaded files. Extract features or generate transformer patches for deep learning applications.
|
309 |
+
|
310 |
+
**Features:**
|
311 |
+
- 📊 **Basic Features**: Waveform, Mel Spectrogram, MFCC, Spectral Analysis, Tempo Detection
|
312 |
+
- 🎼 **Chroma Features**: Harmonic content analysis with multiple extraction methods
|
313 |
+
- 🧩 **Transformer Patches**: Fixed-duration patches for deep learning
|
314 |
+
|
315 |
+
**Requirements**: Dependencies are automatically installed in Hugging Face Spaces via `requirements.txt`.
|
316 |
+
""")
|
317 |
+
|
318 |
+
with gr.Row():
|
319 |
+
with gr.Column(scale=1):
|
320 |
+
gr.Markdown("### 📁 Audio Input")
|
321 |
+
with gr.Group():
|
322 |
+
gr.Markdown("**Download from YouTube** (Supported formats: MP3, WAV, etc.)")
|
323 |
+
youtube_url = gr.Textbox(
|
324 |
+
label="YouTube URL",
|
325 |
+
placeholder="https://www.youtube.com/watch?v=...",
|
326 |
+
)
|
327 |
+
download_btn = gr.Button("📥 Download Audio", variant="primary")
|
328 |
+
download_status = gr.Textbox(label="Download Status", interactive=False)
|
329 |
+
|
330 |
+
with gr.Group():
|
331 |
+
gr.Markdown("**Or upload audio file** (Supported formats: MP3, WAV, FLAC, etc.)")
|
332 |
+
audio_file = gr.Audio(
|
333 |
+
label="Upload Audio File",
|
334 |
+
type="filepath",
|
335 |
+
)
|
336 |
+
|
337 |
+
with gr.Column(scale=2):
|
338 |
+
gr.Markdown("### 🔍 Analysis Results")
|
339 |
+
with gr.Tabs():
|
340 |
+
with gr.Tab("📊 Basic Features"):
|
341 |
+
basic_plot = gr.Image(label="Feature Visualizations")
|
342 |
+
basic_summary = gr.Markdown(label="Feature Summary")
|
343 |
+
basic_btn = gr.Button("🔍 Analyze Basic Features", variant="secondary")
|
344 |
+
|
345 |
+
with gr.Tab("🎼 Chroma Features"):
|
346 |
+
chroma_plot = gr.Image(label="Chroma Visualizations")
|
347 |
+
chroma_summary = gr.Markdown(label="Chroma Summary")
|
348 |
+
chroma_btn = gr.Button("🎼 Analyze Chroma Features", variant="secondary")
|
349 |
+
|
350 |
+
with gr.Tab("🧩 Transformer Patches"):
|
351 |
+
with gr.Row():
|
352 |
+
patch_duration = gr.Slider(
|
353 |
+
label="Patch Duration (seconds)",
|
354 |
+
minimum=1.0, maximum=10.0, value=5.0, step=0.5,
|
355 |
+
)
|
356 |
+
hop_duration = gr.Slider(
|
357 |
+
label="Hop Duration (seconds)",
|
358 |
+
minimum=0.1, maximum=5.0, value=1.0, step=0.1,
|
359 |
+
)
|
360 |
+
patches_plot = gr.Image(label="Generated Patches")
|
361 |
+
patches_summary = gr.Markdown(label="Patch Summary")
|
362 |
+
patches_btn = gr.Button("🧩 Generate Patches", variant="secondary")
|
363 |
+
|
364 |
+
error_output = gr.Textbox(label="Error Messages", interactive=False)
|
365 |
+
|
366 |
+
gr.Markdown("""
|
367 |
+
### ℹ️ Usage Tips
|
368 |
+
- **Processing Limits**: 60s for basic features, 30s for chroma features for fast response
|
369 |
+
- **YouTube Downloads**: Ensure URLs are valid and respect YouTube's terms of service
|
370 |
+
- **Visualizations**: High-quality, suitable for research and education
|
371 |
+
- **Storage**: Temporary files are cleaned up when the interface closes
|
372 |
+
- **Support**: For issues, check the [GitHub repository](https://github.com/your-repo)
|
373 |
+
""")
|
374 |
+
|
375 |
+
# Event handlers
|
376 |
+
download_btn.click(
|
377 |
+
fn=analyzer.download_youtube_audio,
|
378 |
+
inputs=[youtube_url],
|
379 |
+
outputs=[audio_file, download_status]
|
380 |
+
)
|
381 |
+
|
382 |
+
basic_btn.click(
|
383 |
+
fn=analyzer.extract_basic_features,
|
384 |
+
inputs=[audio_file],
|
385 |
+
outputs=[basic_plot, basic_summary, error_output]
|
386 |
+
)
|
387 |
+
|
388 |
+
chroma_btn.click(
|
389 |
+
fn=analyzer.extract_chroma_features,
|
390 |
+
inputs=[audio_file],
|
391 |
+
outputs=[chroma_plot, chroma_summary, error_output]
|
392 |
+
)
|
393 |
+
|
394 |
+
patches_btn.click(
|
395 |
+
fn=analyzer.generate_patches,
|
396 |
+
inputs=[audio_file, patch_duration, hop_duration],
|
397 |
+
outputs=[patches_plot, patches_summary, error_output]
|
398 |
+
)
|
399 |
+
|
400 |
+
audio_file.change(
|
401 |
+
fn=analyzer.extract_basic_features,
|
402 |
+
inputs=[audio_file],
|
403 |
+
outputs=[basic_plot, basic_summary, error_output]
|
404 |
+
)
|
405 |
+
|
406 |
+
demo.unload(fn=analyzer.cleanup)
|
407 |
+
|
408 |
+
|
409 |
+
|
410 |
+
|
411 |
+
|
412 |
+
|
413 |
+
|
414 |
+
|
415 |
+
|
416 |
+
|
417 |
+
|
418 |
+
|
419 |
+
|
420 |
+
|
421 |
+
|
422 |
+
|
423 |
+
|
424 |
+
|
425 |
+
return demo
|
426 |
+
|
427 |
+
|
428 |
+
|
429 |
+
|
430 |
+
|
431 |
+
|
432 |
+
|
433 |
+
|
434 |
+
|
435 |
+
|
436 |
+
|
437 |
+
|
438 |
+
|
439 |
+
|
440 |
+
|
441 |
+
|
442 |
+
|
443 |
+
|
444 |
+
|
445 |
+
|
446 |
+
|
447 |
+
|
448 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
449 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
450 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
451 |
|
452 |
if __name__ == "__main__":
|
453 |
+
demo = create_gradio_interface()
|
454 |
+
demo.launch()
|