songsee

• You want a fast, scriptable way to generate spectrogram images from MP3 or WAV files without writing code.
• You need multi-panel audio feature visualizations (mel, chroma, HPSS, tempogram, MFCC, etc.) for music analysis, research, or debugging.
• You want to inspect a specific time slice of a long audio file by providing --start and --duration flags.
• You are building a pipeline that pipes audio through stdin and expects an image file as output.

• You need real-time or interactive audio visualization — songsee is a batch CLI tool that writes static image files.
• You need programmatic access to raw feature data (e.g., numpy arrays or JSON) rather than rendered images.

Render multiple acoustic analysis panels in a single grid image using --viz with a comma-separated list. Supported panels include spectrogram, mel, chroma, hpss, selfsim, loudness, tempogram, mfcc, and flux.

Focus on any segment of an audio file using --start and --duration flags. This is useful for inspecting specific moments in long recordings without pre-editing the file.

Accepts audio files directly or via stdin (cat track.mp3 | songsee -), making it easy to integrate into shell pipelines. WAV and MP3 are decoded natively; other formats are handled via ffmpeg if available.

Choose from multiple color palettes (classic, magma, inferno, viridis, gray) and control output dimensions with --width / --height. Output format can be jpg or png.

Fine-tune the analysis with --window and --hop for FFT settings, and --min-freq / --max-freq to constrain the frequency range displayed.

Producers can run songsee track.mp3 to instantly generate a spectrogram of a mix, visually inspecting frequency balance, clipping, or noise before export.

Researchers can generate a full 9-panel visualization covering mel, chroma, HPSS, tempogram, MFCC, and more in one command, providing a comprehensive acoustic fingerprint of a recording.

Using stdin support, songsee can be embedded in CI or batch processing pipelines to automatically generate spectrogram thumbnails or diagnostic images for audio datasets.

Audio engineers can use --start and --duration to isolate and visualize a specific section of a track — such as the first chorus or a suspected artifact — without editing the source file.