report_generator.py

import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
import librosa.display
import numpy as np


def create_report(audio_data, output_path):
    """
    Create a complete forensic PNG report.
    Synthetic detection is informational only.
    """

    plt.style.use("default")

    fig = plt.figure(figsize=(22, 16))
    fig.patch.set_facecolor("white")

    fig.suptitle(
        f"AUDIO FORENSIC ANALYSIS REPORT\n{audio_data['filename']}",
        fontsize=20,
        fontweight="bold",
        y=0.97
    )

    gs = gridspec.GridSpec(
        5, 4,
        figure=fig,
        hspace=0.45,
        wspace=0.4,
        height_ratios=[1.5, 1, 0.8, 0.9, 0.7],
        left=0.05,
        right=0.95,
        top=0.92,
        bottom=0.05
    )

    # ============================================================
    # 1. SPECTROGRAM PANEL
    # ============================================================

    ax_spec = fig.add_subplot(gs[0, :])

    S_db = audio_data["spectral"]["S_db"]
    sr = audio_data["info"]["samplerate"]
    hop = audio_data["spectral"]["hop_length"]

    img = librosa.display.specshow(
        S_db,
        sr=sr,
        hop_length=hop,
        y_axis="hz",
        x_axis="time",
        cmap="viridis",
        ax=ax_spec,
        vmin=-80,
        vmax=0
    )

    ax_spec.set_title("Spectrogram", fontsize=14, fontweight="bold", pad=10)
    ax_spec.grid(True, alpha=0.3, linestyle="--")
    cbar = plt.colorbar(img, ax=ax_spec, pad=0.01)
    cbar.set_label("Magnitude (dB)", fontsize=10, fontweight="bold")

    # ============================================================
    # 2. FILE INFORMATION PANEL
    # ============================================================

    ax_info = fig.add_subplot(gs[1, 0:2])
    ax_info.axis("off")

    info = audio_data["info"]
    t = audio_data["time_stats"]

    lines = [
        "FILE INFORMATION",
        "─" * 50,
        f"Sample Rate:     {info['samplerate']:,} Hz",
        f"Channels:        {info['channels']}",
        f"Duration:        {info['duration']:.2f} sec",
        f"Format:          {info['format']} ({info['subtype']})",
        f"Frames:          {info['frames']:,}",
        "",
        "TIME ANALYSIS",
        "─" * 50,
        f"Peak:            {t['peak_db']:.2f} dBFS ({t['peak']:.6f})",
        f"RMS:             {t['rms_db']:.2f} dBFS ({t['rms']:.6f})",
        f"Crest Factor:    {t['crest_factor_db']:.2f} dB",
        f"Noise Floor:     {t['noise_floor']:.6f}",
        f"Est. SNR:        {t['snr_db']:.1f} dB",
        f"Zero Cross Rate: {t['zero_crossing_rate']:.4f}",
    ]

    if audio_data.get("lufs") is not None:
        lines += [
            "",
            "LOUDNESS",
            "─" * 50,
            f"Integrated LUFS: {audio_data['lufs']:.2f}"
        ]

    ax_info.text(
        0.05, 0.95,
        "\n".join(lines),
        fontsize=10.8,
        va="top",
        family="monospace",
        bbox=dict(
            boxstyle="round,pad=1",
            facecolor="#E8F4F8",
            edgecolor="#0077BE",
            linewidth=2
        )
    )

    # ============================================================
    # 3. SPECTRAL STATS PANEL
    # ============================================================

    ax_specstats = fig.add_subplot(gs[1, 2:4])
    ax_specstats.axis("off")

    spec = audio_data["spectral"]
    e = spec["energy_distribution"]

    text = [
        "SPECTRAL ANALYSIS",
        "─" * 50,
        f"Centroid:        {spec['spectral_centroid']:.1f} Hz",
        f"Bandwidth:       {spec['spectral_bandwidth']:.1f} Hz",
        f"Flatness:        {spec['spectral_flatness']:.4f}",
        f"Rolloff Mean:    {spec['spectral_rolloff']:.1f} Hz",
        "",
        "ROLLOFF POINTS",
        "─" * 50,
        f"85% Energy:      {spec['rolloff_85pct']:.1f} Hz",
        f"95% Energy:      {spec['rolloff_95pct']:.1f} Hz",
        f"Highest -60 dB:  {spec['highest_freq_minus60db']:.1f} Hz",
        "",
        "ENERGY DISTRIBUTION",
        "─" * 50,
        f"< 100 Hz:        {e['below_100hz']:.2f}%",
        f"100–500 Hz:      {e['100_500hz']:.2f}%",
        f"500–2k Hz:       {e['500_2khz']:.2f}%",
        f"2k–8k Hz:        {e['2k_8khz']:.2f}%",
        f"8k–12k Hz:       {e['8k_12khz']:.2f}%",
        f"12k–16k Hz:      {e['12k_16khz']:.2f}%",
        f"> 16k Hz:        {e['above_16khz']:.2f}%",
    ]

    ax_specstats.text(
        0.05, 0.95,
        "\n".join(text),
        fontsize=10.8,
        va="top",
        family="monospace",
        bbox=dict(
            boxstyle="round,pad=1",
            facecolor="#FFF4E6",
            edgecolor="#FF8C00",
            linewidth=2
        )
    )

    # ============================================================
    # 4. ENERGY BAR CHART
    # ============================================================

    ax_bar = fig.add_subplot(gs[2, :])

    bands = [
        "<100Hz", "100–500Hz", "500–2kHz",
        "2k–8kHz", "8k–12kHz", "12k–16kHz", ">16kHz"
    ]

    vals = [
        e["below_100hz"], e["100_500hz"], e["500_2khz"],
        e["2k_8khz"], e["8k_12khz"], e["12k_16khz"], e["above_16khz"]
    ]

    colors = ["#2C3E50", "#E74C3C", "#E67E22",
              "#F39C12", "#2ECC71", "#3498DB", "#9B59B6"]

    bars = ax_bar.bar(bands, vals, color=colors, edgecolor="black", alpha=0.85)

    ax_bar.set_ylabel("Energy (%)", fontsize=12, fontweight="bold")
    ax_bar.grid(axis="y", alpha=0.35, linestyle="--")

    for b, v in zip(bars, vals):
        ax_bar.text(b.get_x() + b.get_width()/2, v + 0.3, f"{v:.2f}%", ha="center", fontsize=10)

    # ============================================================
    # 5. ISSUES PANEL
    # ============================================================

    ax_issues = fig.add_subplot(gs[3, 0:3])
    ax_issues.axis("off")

    issues = audio_data["issues"]

    issue_lines = ["DETECTED ISSUES", "═" * 80]

    if not issues:
        issue_lines.append("✅ No significant issues detected.")
    else:
        icons = {
            "CRITICAL": "🔴",
            "HIGH": "🟠",
            "MEDIUM": "🟡",
            "LOW": "🟢"
        }
        for issue, sev, desc in issues:
            issue_lines.append(f"{icons.get(sev,'⚪')} [{sev}] {issue}")
            issue_lines.append(f"   → {desc}")

    if spec["spectral_notches"]:
        issue_lines += [
            "",
            f"🎵 Spectral Notches: {len(spec['spectral_notches'])}",
        ]
        for i, n in enumerate(spec["spectral_notches"][:5], 1):
            issue_lines.append(f"  {i}. {n['freq']:.1f} Hz (Depth {n['depth_db']:.1f} dB)")

    ax_issues.text(
        0.05, 0.95,
        "\n".join(issue_lines),
        fontsize=10.8,
        va="top",
        family="monospace",
        bbox=dict(
            boxstyle="round,pad=1.2",
            facecolor="#FFE6E6",
            edgecolor="#DC143C",
            linewidth=2
        )
    )

    # ============================================================
    # 6. QUALITY SCORE PANEL + SYNTHETIC BLOCK
    # ============================================================

    ax_score = fig.add_subplot(gs[3, 3])
    ax_score.axis("off")

    s = audio_data["score"]
    syn = audio_data["synthetic"]

    score_lines = [
        "QUALITY ASSESSMENT",
        "═" * 28,
        f"SCORE: {s['score']}/100",
        f"GRADE: {s['grade']}",
        f"QUALITY: {s['quality']}",
        "",
        "RECOMMENDATION:",
        s["recommendation"],
        "",
        "CLEANLINESS SCORE:",
        f"{s['cleanliness_score']}/100",
        "",
        "PROCESSING SEVERITY:",
        f"{s['processing_severity']}",
        "",
        "ISSUE SUMMARY",
        "─" * 28,
        f"Critical: {s['critical']}",
        f"High:     {s['high']}",
        f"Medium:   {s['medium']}",
        f"Low:      {s['low']}",
    ]

    score_lines += [
        "",
        "━━━━━━━━━━━━━━━━━━━━━━━",
        "     SYNTHETIC VOICE",
        "━━━━━━━━━━━━━━━━━━━━━━━",
        f"Probability : {syn['synthetic_probability']:.2f}",
        f"Label       : {syn['synthetic_label']}",
        "━━━━━━━━━━━━━━━━━━━━━━━",
        "",
        f"Generated: {audio_data['timestamp']}"
    ]

    ax_score.text(
        0.5, 0.5,
        "\n".join(score_lines),
        fontsize=11,
        ha="center",
        va="center",
        family="monospace",
        bbox=dict(
            boxstyle="round,pad=1.4",
            facecolor=s["color"],
            edgecolor="black",
            linewidth=3,
            alpha=0.70
        )
    )

    # ============================================================
    # SAVE REPORT
    # ============================================================

    plt.savefig(output_path, dpi=300, bbox_inches="tight")
    plt.close()

    return output_path