app.py

#!/usr/bin/env python3
"""
Audio Visualizer — Gradio Web Interface for Hugging Face Spaces
Upload audio, get 22 visualizations + a zip download ready for Claude.
"""

import matplotlib
matplotlib.use('Agg')  # MUST be before any pyplot/librosa import

import gc
import shutil
import tempfile
from pathlib import Path

import gradio as gr
import librosa
import matplotlib.pyplot as plt
import numpy as np

import audio_visualizer

# All 22 visualization functions in order (mirrors the GUI list)
VISUALIZATIONS = [
    ("Waveform", audio_visualizer.plot_waveform),
    ("Volume Envelope", audio_visualizer.plot_waveform_envelope),
    ("Spectrogram", audio_visualizer.plot_spectrogram),
    ("Mel Spectrogram", audio_visualizer.plot_mel_spectrogram),
    ("Chromagram", audio_visualizer.plot_chromagram),
    ("Tonnetz", audio_visualizer.plot_tonnetz),
    ("Spectral Centroid", audio_visualizer.plot_spectral_centroid),
    ("Spectral Bandwidth", audio_visualizer.plot_spectral_bandwidth),
    ("Spectral Rolloff", audio_visualizer.plot_spectral_rolloff),
    ("RMS Energy", audio_visualizer.plot_rms_energy),
    ("Zero Crossing Rate", audio_visualizer.plot_zero_crossing_rate),
    ("Onset Strength", audio_visualizer.plot_onset_strength),
    ("Beat Tracking", audio_visualizer.plot_beat_track),
    ("Tempogram", audio_visualizer.plot_tempogram),
    ("MFCCs", audio_visualizer.plot_mfcc),
    ("Spectral Contrast", audio_visualizer.plot_spectral_contrast),
    ("Harmonic/Percussive", audio_visualizer.plot_harmonic_percussive),
    ("Frequency Bands", audio_visualizer.plot_frequency_bands),
    ("Dynamic Range", audio_visualizer.plot_dynamic_range),
    ("Spectral Flatness", audio_visualizer.plot_spectral_flatness),
    ("Combined Dashboard", audio_visualizer.plot_combined_dashboard),
    ("3D Spectrogram", audio_visualizer.plot_3d_spectrogram),
]

DPI_OPTIONS = {
    "Normal (150 DPI)": 150,
    "High (200 DPI)": 200,
    "Ultra (300 DPI)": 300,
}


def generate_visualizations(audio_path, quality, progress=gr.Progress()):
    """Generate all 22 visualizations and return gallery images + zip file."""
    if audio_path is None:
        raise gr.Error("Please upload an audio file first.")

    # Set DPI
    audio_visualizer.FIGURE_DPI = DPI_OPTIONS.get(quality, 150)

    # Load audio
    progress(0, desc="Loading audio...")
    y, sr = audio_visualizer.load_audio(audio_path)
    duration = librosa.get_duration(y=y, sr=sr)
    audio_file = Path(audio_path)
    title = audio_file.stem

    # Create temp output directory
    output_tmp = tempfile.mkdtemp(prefix="avis_output_")
    output_dir = Path(output_tmp)

    # Generate each visualization
    total = len(VISUALIZATIONS)
    image_paths = []

    for i, (name, func) in enumerate(VISUALIZATIONS):
        progress((i) / total, desc=f"Generating: {name} ({i + 1}/{total})...")

        if func == audio_visualizer.plot_combined_dashboard:
            func(y, sr, output_dir, base_path=audio_file)
        else:
            func(y, sr, output_dir)

        plt.close('all')
        gc.collect()

    # Create visualization guide
    progress(0.95, desc="Creating visualization guide...")
    tempo, _ = librosa.beat.beat_track(y=y, sr=sr)
    audio_visualizer.create_visualization_guide(output_dir, duration, tempo, title)

    # Collect all PNG paths (sorted by filename for correct order)
    image_paths = sorted(output_dir.glob("*.png"))

    # Create zip file
    progress(0.98, desc="Creating zip archive...")
    zip_tmp = tempfile.mkdtemp(prefix="avis_zip_")
    zip_base = Path(zip_tmp) / f"{title}_visualizations"
    zip_path = shutil.make_archive(str(zip_base), 'zip', output_dir)

    progress(1.0, desc="Done!")

    return image_paths, zip_path


# --- Build the Gradio interface ---

with gr.Blocks(
    title="Audio Visualizer",
    theme=gr.themes.Soft(),
) as demo:
    gr.Markdown(
        """
        # Audio Visualizer — Let Claude Hear Your Music
        Upload any audio file to generate **22 visualizations** that translate sound into sight.
        Download the zip and share it with Claude to let AI "listen" to your music.
        """
    )

    with gr.Row():
        with gr.Column(scale=1):
            audio_input = gr.Audio(
                type="filepath",
                label="Upload Audio File",
            )
            quality_radio = gr.Radio(
                choices=list(DPI_OPTIONS.keys()),
                value="Normal (150 DPI)",
                label="Quality",
            )
            generate_btn = gr.Button("Generate Visualizations", variant="primary")
        with gr.Column(scale=1):
            gr.Markdown(
                """
                ### How it works
                1. **Upload** an MP3, WAV, FLAC, OGG, or other audio file
                2. **Choose quality** — higher DPI = sharper images but slower
                3. **Click Generate** and wait for all 22 visualizations
                4. **Download the zip** and upload it to a Claude conversation

                Claude can analyze these images to describe the music's rhythm,
                melody, dynamics, and texture — even though it can't hear the
                audio directly.
                """
            )

    gallery = gr.Gallery(
        label="Visualizations",
        columns=4,
        object_fit="contain",
        height="auto",
    )

    zip_download = gr.File(label="Download All (Zip)")

    generate_btn.click(
        fn=generate_visualizations,
        inputs=[audio_input, quality_radio],
        outputs=[gallery, zip_download],
    )

if __name__ == "__main__":
    demo.launch()