Sound Visualizer

A small Windows desktop audio visualizer using NumPy, PyQt6, and SoundCard/WASAPI.

Setup

cd C:\Users\Sopht\coding\sound_visualizer
python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt
python main.py

The first run uses the Windows default speaker loopback. To use a microphone, edit the JSON file at:

%APPDATA%\SoundVisualizer\config.json

Create it from the defaults with:

python config.py

Set audio.source to input for the default microphone, or set audio.device to a device id printed by:

python audio_capture.py

Configuration

The configuration exposes independent output/input sources, FFT sizes, shared frequency bounds and smoothing, bar count, colors, opacity, frame rate, visualization mode, line style, and data overlays.

• visual.mode: bars or line
• visual.orientation: flat or circle
• visual.line_style: trace or organic
• visual.radius_ratio: circle radius as a fraction of the smaller window dimension
• visual.left_color and visual.right_color: independent stereo colors
• general.frequency_min and general.frequency_max: shared frequency range for both streams
• general.smoothing and general.decay: shared FFT response controls
• general.binning_method: default preserves the existing grouping; logarithmic uses expanding high-frequency bins
• data.marker_start and data.marker_end: marker span from 0 to 1
• data.marker_opacity, data.marker_color, data.data_opacity, data.data_font, and data.data_size_factor: overlay styling
• data.show_output_frequency, data.show_output_bpm, data.show_input_frequency, and data.show_input_bpm: independent stream readouts

The tray icon is shown in the Windows notification area. Its Settings action opens a separate live configuration window with color pickers and expandable input/output device selectors. Changes are saved immediately. When click-through is disabled, drag anywhere to move the visualizer and drag the lower-right grip to resize it.

Output and input have independent capture and magnitude controls:

• analysis.output_magnitude_mode: normalized or scaled
• analysis.output_magnitude_scale: used by scaled
• analysis.input_magnitude_mode: normalized or scaled
• analysis.input_magnitude_scale: used by scaled

Each device selector has Default and None (disabled). Selecting None removes that stream's visualization and data readouts. Existing configurations are migrated when loaded.

Windows startup

Startup registration is available from Python:

python -c "from startup import set_startup; set_startup(True)"

Disable it with set_startup(False). This uses the current user's Run key and does not require administrator rights.

Troubleshooting

Run python capture_probe.py to enumerate input and loopback devices and confirm that audio blocks are arriving. If the selected device is unavailable, the visualizer remains open and renders a silent state while the tray can be used to quit or reload configuration.

The desktop layer is currently a Windows-compatible fallback mode for a non-topmost transparent widget. Reliable attachment behind desktop icons will be implemented as a separate Windows shell integration step.

The default capture is stereo for both output loopback and input. Left and right channels are analyzed independently and drawn on top of each other. In flat orientation output grows upward and input grows downward; in circle orientation output grows outward and input grows inward. A few inactive frequency channels can be normal when an FFT bin does not land inside a narrow logarithmic band; the analyzer now samples the spectrum at logarithmic frequency centers instead, which keeps those bands continuous. The lowest and highest bands can still be quiet when they fall outside the useful audio/device range.