songsee

Why it's ranked here

The appeal is practical breadth in a small Go tool. Nine analysis views cover frequency, pitch class, timbre, tempo, dynamics, structure, and spectral change. Shared analysis data avoids needless recomputation, percentile normalization improves panel readability, and tests exercise every rendering mode.

What's good

It installs through Homebrew, the Go toolchain, or Docker. The container includes FFmpeg, which makes batch and server jobs easier to reproduce. Users can combine views, choose six palettes, crop by time, restrict frequency ranges, tune FFT settings, and control image dimensions.

Tradeoffs

The documented interface produces static images, not live or interactive visualization. Audio is reduced to mono, so stereo placement disappears. WAV and MP3 decode internally, but other formats require an external FFmpeg executable unless the Docker image is used. Output is limited to PNG and JPEG.

How to use it well

Use it for quick inspection, batch-generated audio reports, dataset previews, or comparing several feature views of the same passage. Start with the default spectrogram, then add chroma, loudness, flux, or self-similarity when the question demands them. Choose another tool for playback, editing, stereo analysis, or interactive exploration.

Observed

Primary language

Go, with the module targeting Go 1.25

Install surfaces

Homebrew formula, Go install command, and Docker build

Interface

Command-line application

Audio input

Native WAV and MP3 decoding, with FFmpeg fallback for other formats

Image output

PNG or JPEG with configurable width and height

Visualization surface

Nine modes and six color palettes, with multiple modes composable into one grid

Tests

Repository text includes FFT tests and visualization tests covering all nine rendering modes