Guide · Agent skill
How to use the songsee Audio Visualizer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The songsee skill gives an AI agent the ability to run the songsee command-line tool, which turns audio files into pictures such as spectrograms. This guide is for anyone who has never used an agent skill before, and by the end you will know how to install the skill, ask for a visualization in plain language, and understand what the output image can and cannot tell you.
songsee Audio Visualizerspinabot/brigade · MIT Download the skill Quick start
- 1Download the skill ZIP. Get the songsee skill ZIP from this page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings, Capabilities, Skills in the Claude apps.
- 3Install the songsee tool. Install the songsee command-line tool, for example with brew install steipete/tap/songsee.
- 4Check the command runs. Run songsee on a sample file in your terminal to confirm it works.
- 5Ask the agent for a visualization. Name your audio file and describe the picture you want, such as a spectrogram or a multi-panel grid.
- 6Review the image. Open the jpg or png output and interpret it by eye.
What the songsee skill makes the agent do
A spectrogram shows how the frequencies in a recording change over time. The songsee skill teaches the agent to make these images from an audio file using the songsee command-line tool.
songsee can also draw several analysis panels, such as tempo, chroma and loudness, arranged in one grid. The agent can pick a time slice, a colour style or a frequency range, and produce an image file for you to view or use.
- Spectrogram of a whole track or a short slice
- Multi-panel grids with features such as mel, chroma, hpss, selfsim, loudness, tempogram, mfcc and flux
- Choice of colour palettes: classic, magma, inferno, viridis or gray
- Output as jpg or png
Install the songsee skill and its tool
Download the skill ZIP from our page and unzip it into ~/.claude/skills/ for your user account, or into a project's .claude/skills/ folder to limit it to that project. Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
The skill also needs the songsee command-line tool. The skill's metadata lists a Homebrew install using the formula steipete/tap/songsee. Check that the songsee command runs in your terminal before asking the agent to use it.
- Skill file included: SKILL.md
- Tool install label in the source: Install songsee (brew)
brew install steipete/tap/songseeAsk for a visualization in plain language
You do not need to remember flags. Describe what you want, name the audio file, and say what kind of picture you are after. The agent chooses the matching options.
Useful details to include are the file, the time window you care about, and whether you want one panel or several. If you do not give a time window, the agent will use the whole file.
- "Make a spectrogram of track.mp3."
- "Show an 8-second slice starting at 12.5 seconds from track.mp3 as a jpg."
- "Draw spectrogram, chroma and tempogram panels for track.mp3 in one image."
Examples of the commands the skill uses
These are the quick-start patterns from the skill. You can run them yourself, or let the agent run them for you.
The --viz option takes a list, either comma-separated or repeated. When you give more than one panel, songsee renders a grid. The --start and --duration options cut a time slice, and -o sets the output file.
- --style sets the palette (classic, magma, inferno, viridis, gray)
- --width and --height set the output size
- --window and --hop adjust the FFT settings
- --min-freq and --max-freq set the frequency range
- --format selects jpg or png
songsee track.mp3
songsee track.mp3 --viz spectrogram,mel,chroma,hpss,selfsim,loudness,tempogram,mfcc,flux
songsee track.mp3 --start 12.5 --duration 8 -o slice.jpg
cat track.mp3 | songsee - --format png -o out.pngWhat it needs and what to expect
WAV and MP3 files are decoded natively by songsee. For other formats, songsee uses ffmpeg, but only if ffmpeg is installed on your system. ffmpeg is optional if you only work with WAV and MP3.
The output is a picture. Read it by eye. The skill does not produce exact BPM, key or other numeric analysis, and it does not edit or convert the source audio.
- Required: the songsee command-line tool
- Optional: ffmpeg, only for formats other than WAV and MP3
- Runs scripts: yes
- Uses network: no
Safety notes on scripts and network
This skill runs scripts, so the agent will execute the songsee command on your machine. Review the commands it proposes if you are working with sensitive files, as you would with any tool the agent can run.
According to the listing, the skill does not use the network. The source file itself also states that songsee does not modify the source audio file.
- Runs scripts: yes, the songsee command is executed locally
- Uses network: no
- Source audio is not modified by songsee
Tips for better results
Start with the default spectrogram to get a feel for the track, then add panels one at a time. Use a time slice when a full track makes the image too dense to read.
The --window and --hop settings change how fine the time and frequency detail looks, so try a few values if the default is not clear. The gray style can be easier to read for printing.
- Use --start and --duration to zoom into one section
- Add --min-freq and --max-freq to focus on a frequency band
- Use png for lossless output and jpg for smaller files
What people use it for
- Musicians and producers checking the structure of a track, such as where the energy builds and drops.
- Audio students learning how spectrograms, tempo and chroma panels look for different sounds.
- Developers in audio analysis pipelines who want a quick visual check of their files.
- Making a time-sliced image of one section of a longer recording.
Frequently asked questions
How do I install the songsee skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings, Capabilities, Skills.
Do I need to install anything else?
Yes, the skill needs the songsee command-line tool. The listed install method is Homebrew with the formula steipete/tap/songsee. ffmpeg is optional and only needed for formats other than WAV and MP3.
Is songsee free to use?
The source lists the skill under the MIT licence. This page does not state pricing for songsee beyond that licence, so check the project's repository for its terms.
Which audio formats work?
WAV and MP3 are decoded natively. Other formats work if ffmpeg is installed, because songsee then uses it to read them.
Does the skill need internet access?
No. The listing says the skill uses no network. It does run scripts, so the agent executes the songsee command locally.
Can songsee tell me the exact BPM or key of a track?
No. The skill produces images to interpret by eye, not numeric analysis such as exact BPM or key. For those numbers you would need a different tool.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the songsee skill if you want to see the structure of a track or check audio visually, and you are comfortable with an agent running a local command-line tool. Pick something else if you need to edit or convert audio, or if you need exact numeric results like BPM or key.
Download the skill