Generate spectrograms and audio feature visualizations from audio files via CLI.
npx skills add https://github.com/MarbleSodas/Mavis --skill songsee-marblesodas
Please help me install this Agent Skill. Skill: songsee Source: https://github.com/MarbleSodas/Mavis/tree/main/skills/media/songsee Command: npx skills add https://github.com/MarbleSodas/Mavis --skill songsee-marblesodas
songsee removes the manual work of visually inspecting audio by generating spectrograms and common audio features so you can quickly understand what’s happening in a track.
Use case: Compare two versions of a song by generating matching grids (e.g., spectrogram, mel, MFCC, flux) to spot differences in pitch content, onset activity, and timbral changes.
Run songsee on an MP3 file and save a multi-panel visualization grid to an image with spectrogram, mel, chroma, mfcc, and flux.
High-intent search queries and answers about installing and using this skill.
You can generate a spectrogram by running the songsee command on an MP3 or WAV file to produce PNG or JPG image outputs. It natively processes WAV and MP3 formats, while using ffmpeg to handle other audio file types.
Yes, you can visualize MFCC and chroma features by using the --viz option to render multi-panel outputs. This includes spectral flux, tempogram, self-similarity, and loudness grids for comprehensive music analysis.
FFmpeg is only required for audio formats beyond native WAV and MP3 files. The tool processes standard WAV and MP3 inputs directly without additional dependencies, while relying on ffmpeg to decode and convert other formats.
You can extract a spectrogram for a specific time slice by running the CLI command with targeted time parameters for focused debugging. This allows you to isolate and visually inspect frequency content and temporal events within a specific segment.
The best way to compare two versions is to generate matching multi-panel visualization grids for both tracks. You can use the CLI to output consistent images showing spectrogram, mel, MFCC, and spectral flux data to spot differences in pitch and onset activity.
A limitation of this CLI visualization method is that it outputs static PNG or JPG image artifacts rather than interactive plots. It is designed for consistent documentation and visual inspection, not for real-time signal manipulation or dynamic onset investigation.
TitoPrauseeGenerate spectrogram and audio feature visualizations from audio files.
AlvaroBianoGenerate spectrogram, mel, chroma, hpss, mfcc visualizations from audio files to PNG/JPG using songsee CLI.
StarlitnightlyGenerate publication-quality plots with matplotlib and seaborn.
dawsonblockGenerate spectrograms and multi-panel audio feature visualizations from files.
