Audio Spectrum Analyzer
This browser-local analyzer decodes your audio and measures the changing frequency content of a selected 30-second segment. A Hann-window FFT produces a logarithmic-frequency spectrogram; the inspected frame reports its dominant frequency, up to five local peaks and RMS level. Choose the FFT window, hop and stereo channel to see how time and frequency resolution trade off. These are measurements of browser-decoded PCM, not musical note recognition or mastering advice.
Key features
- Analyze a selected range of browser-decodable WAV, MP3, OGG, WebM, M4A or FLAC audio without uploading it.
- Choose 512–4096 sample Hann windows, 25/50/100% hop and mixed, left or right channels.
- Inspect a labeled logarithmic spectrogram, window-center time, dominant frequency, local peaks and RMS dBFS.
- Save a chart PNG, per-window peak/RMS CSV or the selected frame's frequency-bin CSV.
- Use the generated 440-to-880 Hz WAV example through the same browser decode and analysis path.
How to use
- Choose a local audio file up to 12 MiB or load the generated WAV example.
- Set the start/end seconds, FFT window, hop percentage and stereo channel.
- Analyze the segment, then move the frame slider to inspect frequency peaks over time.
- Download the spectrogram PNG and whichever CSV measurements you need.
Use cases
- Check whether a recorded tone switches from one frequency to another.
- Compare the left and right channels of a stereo recording for different tones.
- Review a short noise or music segment's visible frequency distribution before editing elsewhere.
Frequently asked questions
What do FFT size and hop change?
A larger FFT window produces narrower frequency bins but averages a longer moment; a shorter hop places more overlapping windows on the time axis. Changing the hop does not add frequency resolution. All windows use a Hann taper.
Why can a peak frequency differ slightly from the source tone?
FFT bins are separated by sample rate divided by window size. The displayed peak uses a neighboring-bin interpolation estimate, and short windows, harmonics, noise and browser resampling can move it. The number is not a calibrated instrument reading.
What does dBFS mean here?
Peak values are approximate one-sided FFT amplitudes relative to full-scale PCM, with a Hann coherent-gain correction; RMS is computed in the original window. Results are clamped at -100 dBFS for display. They are not sound-pressure level or loudness units.
Which files and segments are allowed?
The file must be at most 12 MiB and a recognizable audio container. Decoded audio must be 0.25–90 seconds, mono/stereo, 8–96 kHz and at most eight million frames. Each analysis covers at most 30 seconds and 3,000 FFT windows. Codec support varies by browser.
Does the chart show every frequency?
The PNG draws the bins on a logarithmic scale from 50 Hz to Nyquist, so frequencies below 50 Hz are omitted from the picture. The selected-frame CSV contains all FFT bins from 0 Hz through Nyquist; the timeline CSV contains each window's dominant peak and RMS.
Is my audio uploaded or changed?
No. The browser decodes and analyzes the chosen file locally. Exports contain a chart or measurements, never a modified source file. No audio samples are included in either CSV.
Privacy
Audio is decoded and measured only in browser memory. The page does not send the file to a server API. Clearing the page removes the in-memory PCM and results; downloaded PNG/CSV files remain wherever you save them.
Comments & questions