A modern OpenGL-rendered spectrogram analyzer for music enthusiasts and HiFi listeners
一款面向音乐爱好者与 HiFi 发烧友的现代化频谱分析工具
| English | 中文 |
|---|---|
| OpenGL Spectrogram — fragment shader with real-time colormap switching (10 palettes) | OpenGL 频谱图 — 片段着色器渲染，实时切换 10 种色板 |
| Audio Playback — play/pause with progress slider, synchronized with waveform | 音频播放 — 播放/暂停，进度滑块与波形同步 |
| Progress Slider — dedicated seek bar between spectrogram and X-axis | 播放进度条 — 声谱图与时间轴之间的独立滑块，拖拽定位 |
| Cursor Info — hover shows time, frequency, and dB at cursor position | 光标信息 — 悬停实时显示当前时间、频率、dB 值 |
| Wheel Zoom — scroll to zoom time axis, Shift+scroll for frequency axis, double-click reset | 滚轮缩放 — 滚轮缩放时间轴，Shift+滚轮缩放频率轴，双击重置 |
| Quality Analysis — per-channel clipping (bit-depth-aware), upsampling check (Welch PSD + slope + Gibbs), DR, LUFS, true peak, LRA | 质量分析 — 多声道削波（位深感知）、升频检测（Welch PSD + 斜率 + Gibbs）、动态范围、响度、真峰值 |
| Multi-resolution STFT — standard, multi-band, and phase-reassigned (iZotope RX style) | 多分辨率 STFT — 标准、多频段、相位重分配三种模式 |
| Flexible Y-axis — linear, logarithmic, mel (Slaney piecewise), and bark frequency scales | 灵活 Y 轴 — 线性、对数、mel（Slaney 分段拟合）、bark 四种频率刻度 |
| Waveform Preview — down-sampled envelope aligned with spectrogram | 波形预览 — 降采样包络，与频谱上下对齐 |
| Drag & Drop — WAV, FLAC, MP3, M4A, OGG, AAC, OPUS, APE, AIFF, WMA, TTA | 拖放加载 — 支持 11 种常见音频格式 |
| Bilingual UI — Chinese / English one-click toggle | 双语界面 — 工具栏一键切换中文 / 英文 |
| Batch Analysis — process folders, export to CSV | 批量分析 — 批量处理文件夹，导出 CSV |
| Screenshot Export — save spectrogram view as PNG | 截图导出 — 保存当前频谱视图为 PNG |
| Fast Startup — heavy libraries lazy-loaded in background thread | 快速启动 — 重量级库后台延迟加载，启动秒开 |
| Optimized Analysis — vectorized FFT, deduped true-peak, batched spectrum computation | 优化分析 — 向量化 FFT、去重真峰值、批量频谱计算 |
| Format | Extension |
|---|---|
| FLAC | .flac |
| WAV | .wav |
| MP3 | .mp3 |
| M4A / ALAC | .m4a .mp4 |
| AAC | .aac |
| OGG Vorbis | .ogg |
| OPUS | .opus |
| Monkey's Audio | .ape |
| WMA | .wma |
| True Audio | .tta |
| AIFF | .aiff |
All formats decoded via PyAV (libav). No external dependencies required.
全格式通过 PyAV (libav) 解码，无需外部依赖。
| Metric | Description |
|---|---|
| Clipping | Per-channel flat-top detection (threshold 0.999) — hard/soft classification via bit-depth-aware flatness, single-sample + multi-sample events, channels_affected |
| Hi-freq Cutoff | Welch PSD (nperseg=8192) + multi-factor scoring — noise floor P5, signal ref P90 (2–12 kHz), spectral slope (dB/oct), Gibbs ringing detection |
| Dynamic Range | P95–P10 of segmented frame RMS (TT DR Meter standard) |
| LUFS (I) | EBU R128 integrated loudness |
| LUFS (S) | Short-term loudness (3 s blocks, maximum) |
| LRA | Loudness range (P10–P95 spread) |
| True Peak | 4× oversampling BS.1770-4 |
| Peak / RMS | Global sample peak and RMS level |
git clone https://github.com/Lithivm/Spectra.git
cd Spectra
pip install -r requirements.txt
python main.pyDrag an audio file into the window, or click Open File (打开文件).
拖入音频文件，或点击 打开文件 按钮。Click ▶ (播放) to preview audio.
Download Spectra-v0.2.0-windows.zip from Releases, extract, and run Spectra.exe.
从 Releases 下载 Spectra-v0.2.0-windows.zip，解压后运行 Spectra.exe。
pip install pyinstaller
pyinstaller spectra.spec --noconfirmOutput is in dist/Spectra/. See spectra.spec for packaging details.
输出在 dist/Spectra/，打包配置见 spectra.spec。
| Name | Description |
|---|---|
spectra |
Custom palette with brightness curve (default) |
inferno |
Perceptually uniform, warm |
viridis |
Perceptually uniform, blue-green-yellow |
plasma |
Perceptually uniform, purple-orange-yellow |
magma |
Perceptually uniform, dark-purple-yellow |
hot |
Black → red → yellow |
coolwarm |
Cool blue → warm red (diverging) |
seismic |
Blue → white → red (diverging) |
turbo |
Perceptually uniform rainbow (Google) |
jet |
Classic rainbow |
spectra uses a custom colormap with a dedicated brightness curve (SPECTRA_CURVE); all other palettes are generated from hardcoded color stops (sourced from matplotlib originals) via pure-numpy linear interpolation — no matplotlib dependency. Curve parameters are defined in palette.py and require no changes to any other file to adjust.
Spectra/
├── main.py # 入口薄壳（开发期直跑；装成包后走 `spectra` 命令）
├── docs/ # 文档：specs / agents
├── src/spectra/
│ ├── __main__.py # Entry point + crash logger
│ ├── lang.py # Bilingual (zh/en) toggle
│ ├── assets/ # logo.ico · logo.png · logo_mark.png
│ ├── backend/ # 后端：解码 + 分析
│ │ ├── core.py # AudioAnalyzer facade class
│ │ ├── _state.py # FFTW wisdom + STFT cache (LRU)
│ │ ├── spectrum.py # _SpectrumMixin — STFT, reassigned, mel
│ │ ├── quality.py # _QualityMixin — clipping, LUFS, DR
│ │ ├── loudness.py # R128 K-weighting 原生采样率 + 真峰值（polyphase FIR）
│ │ ├── scales.py # Slaney mel 刻度（CPU 侧唯一实现，shader mel 分支是它的镜像）
│ │ ├── load.py # PyAV multi-format decoder
│ │ ├── metadata.py # Tag extraction via mutagen
│ │ ├── palette.py # Colormap registry + curve presets (zero-dependency)
│ │ └── batch.py # CSV export for batch analysis
│ └── frontend/ # 前端：Qt + OpenGL
│ ├── main_window.py # Main window, toolbar, playback slider, workers
│ ├── spectrogram_widget.py # QOpenGLWidget renderer + axis/colorbar + cursor zoom
│ ├── waveform_widget.py # Waveform envelope
│ ├── metadata_panel.py # File info + quality analysis panel
│ ├── playback_engine.py # Audio playback via sounddevice
│ ├── batch_dialog.py # Batch progress dialog
│ ├── progress_bar.py # 浮在声谱图上的加载进度浮层
│ ├── anim.py # FloatAnim 缓动助手
│ ├── icons.py # 内联 SVG 图标（HiDPI）
│ ├── styles.py # Color tokens for dark theme
│ └── shaders/
│ ├── spectrogram.vert # GLSL vertex shader
│ └── spectrogram.frag # GLSL fragment shader (zoom + curve uniforms)
└── tests/
| Library | Purpose |
|---|---|
| PyQt6 | Desktop UI + OpenGL widget |
| librosa | STFT, spectral features |
| pyFFTW | FFTW bindings (faster FFT) |
| pyloudnorm | EBU R128 loudness |
| PyAV | Audio decoding (libav) |
| sounddevice | Audio playback (PortAudio) |
| mutagen | Metadata extraction |
| scipy | Signal processing |
| numpy | Numerical arrays |
| numba | JIT (librosa dependency) |
If the app crashes, check %USERPROFILE%\.spectra\crash.log for details.
All uncaught exceptions and worker failures are logged there.
如果应用崩溃，查看 %USERPROFILE%\.spectra\crash.log 获取详细错误信息。
MIT
Built with PyQt6, OpenGL, librosa, pyFFTW, matplotlib, and sounddevice.