What audio can I analyze?

Choose a local file up to 50,000,000 bytes. PCM WAV, MP3, M4A, OGG, FLAC and AAC support depends on the browser and codec. Files must contain one mono or stereo audio track at 8,000 to 192,000 samples per second, no longer than 10 minutes or 32 million decoded channel samples. Duration and sample budgets are checked before decoding, and again while samples are collected. Imported audio is never played.

How do FFT size and window affect the spectrum?

Bin spacing is sample rate divided by FFT size. At 48 kHz with 4096 samples it is 11.71875 Hz. Larger windows give finer frequency spacing over a longer time interval. Hann, Hamming, Blackman and rectangular periodic windows trade frequency spreading and leakage. One-sided amplitudes are divided by the sum of window weights; DC and Nyquist are not doubled. A tone between bins can spread across adjacent peaks.

Are the displayed levels sound pressure or loudness?

No. Levels are uncalibrated digital amplitudes expressed in dBFS. A coherent sine with peak sample amplitude 1 is approximately 0 dBFS at its bin. Silence and values below minus 120 dBFS use a finite display floor. Plots clip levels above 0 dBFS visually, while numeric values and CSV retain them. This is not a calibrated sound pressure, equipment or hearing-safety measurement.

How much of the recording is shown in the spectrogram?

The tool samples up to 128 complete FFT windows uniformly across the recording, including the first and last complete windows when more than one window fits. A 300,000-bin cell budget can reduce the number at larger FFT sizes. It does not analyze every instant of a long recording. Each colored frequency pixel shows its maximum included bin level. Time and frequency sliders provide keyboard access to every sampled window and spectrum bin.

How does microphone capture work?

Microphone access starts only when you press Enable microphone and approve the browser request. The latest 10 seconds are kept in a bounded in-memory buffer. Press Stop and analyze microphone to stop every media track and audio node before analysis. Cancel, changing settings, hiding the tab and leaving the page also stop capture. The microphone is never played through the speakers, and no audio is uploaded or saved automatically.

What do channel mean and log scale mean?

Channel mean averages the source sample values, so opposite-polarity stereo signals can cancel. Left / mono and Right inspect individual channels; Right requires stereo audio. Log frequency starts at the first positive FFT bin, because zero has no logarithm. DC remains available in the numeric table and CSV.

What is included in the exports?

CSV includes the selected spectrum and every actual sampled spectrogram bin, with frame number, window start and center times, frequency, normalized amplitude and dBFS. PNG captures the current spectrum, spectrogram, axes and cursors. The copyable report includes the selected frame, normalization details and peak list. All processing occurs in this page and its local worker.

Audio Spectrum Analyzer

Analyze a local audio file or an explicitly captured microphone buffer to inspect its frequency content. Choose FFT size, periodic window, source channel and frequency scale, then move time and frequency cursors through the measured digital samples. Inspect peak bins and the full numeric spectrum, or export every sampled frame and the current plots.

Inspect the frequency content.Local files · explicit microphone capture

Source and FFT settings

Local audio, up to 50 MBNo file selected. Mono/stereo; 8 to 192 kHz; 10 minutes and 32 million decoded channel samples maximum.

Microphone permission alone does not produce a spectrum. Capture audio, then stop and analyze.

FFT sizeWindowPeriodic windows; one-sided amplitude corrected by coherent window gain. Off-bin tones spread across nearby bins.ChannelChannel mean averages sample values and can cancel opposite-polarity stereo signals.Frequency scale

Original example: 48 kHz, 0.5 digital amplitude. At FFT 4096, bin spacing is 11.71875 Hz. Imported audio is never played.

Spectrum and time-frequency view

A spectrum from your audioAnalyze a local file or a captured microphone buffer.
A synthetic tone is available to explore the controls.

Choose local audio or explicitly enable the microphone. Imported audio is never played.

Sample rate: Not analyzed

Bin spacing: Not analyzed

FFT windows: Not analyzed

Selected level: Not analyzed

Accuracy. Uncalibrated digital levels; FFT bin/window limits apply. No equipment or hearing-safety certification.

Common questions

What audio can I analyze?: Choose a local file up to 50,000,000 bytes. PCM WAV, MP3, M4A, OGG, FLAC and AAC support depends on the browser and codec. Files must contain one mono or stereo audio track at 8,000 to 192,000 samples per second, no longer than 10 minutes or 32 million decoded channel samples. Duration and sample budgets are checked before decoding, and again while samples are collected. Imported audio is never played.
How do FFT size and window affect the spectrum?: Bin spacing is sample rate divided by FFT size. At 48 kHz with 4096 samples it is 11.71875 Hz. Larger windows give finer frequency spacing over a longer time interval. Hann, Hamming, Blackman and rectangular periodic windows trade frequency spreading and leakage. One-sided amplitudes are divided by the sum of window weights; DC and Nyquist are not doubled. A tone between bins can spread across adjacent peaks.
Are the displayed levels sound pressure or loudness?: No. Levels are uncalibrated digital amplitudes expressed in dBFS. A coherent sine with peak sample amplitude 1 is approximately 0 dBFS at its bin. Silence and values below minus 120 dBFS use a finite display floor. Plots clip levels above 0 dBFS visually, while numeric values and CSV retain them. This is not a calibrated sound pressure, equipment or hearing-safety measurement.
How much of the recording is shown in the spectrogram?: The tool samples up to 128 complete FFT windows uniformly across the recording, including the first and last complete windows when more than one window fits. A 300,000-bin cell budget can reduce the number at larger FFT sizes. It does not analyze every instant of a long recording. Each colored frequency pixel shows its maximum included bin level. Time and frequency sliders provide keyboard access to every sampled window and spectrum bin.
How does microphone capture work?: Microphone access starts only when you press Enable microphone and approve the browser request. The latest 10 seconds are kept in a bounded in-memory buffer. Press Stop and analyze microphone to stop every media track and audio node before analysis. Cancel, changing settings, hiding the tab and leaving the page also stop capture. The microphone is never played through the speakers, and no audio is uploaded or saved automatically.
What do channel mean and log scale mean?: Channel mean averages the source sample values, so opposite-polarity stereo signals can cancel. Left / mono and Right inspect individual channels; Right requires stereo audio. Log frequency starts at the first positive FFT bin, because zero has no logarithm. DC remains available in the numeric table and CSV.
What is included in the exports?: CSV includes the selected spectrum and every actual sampled spectrogram bin, with frame number, window start and center times, frequency, normalized amplitude and dBFS. PNG captures the current spectrum, spectrogram, axes and cursors. The copyable report includes the selected frame, normalization details and peak list. All processing occurs in this page and its local worker.

Uncalibrated digital levels; FFT bin/window limits apply. No equipment or hearing-safety certification.