Generate band-limited test tones, sweeps and noise from 20 Hz to 20 kHz. Up to eight voices, note-name entry, binaural beats, and WAV export rendered from the same oscillators you hear.
Initializing in your browser…
Cut a region out of any audio file, exact to the sample and measured at zero milliseconds of error. Waveform handles, real keyboard shortcuts, click-free cut fades, and WAV or MP3 output that matches its extension.
Full-track spectrogram with a calibrated dBFS colour scale, labelled Hz axis, and log or linear frequency. Plus four live modes, microphone input, a peak-frequency and note readout, and PNG or WebM export.
Turn audio into a waveform image in five styles, exported as PNG or real vector SVG. Reads every channel, shows true level in dBFS, and redraws correctly at any size.
You need a 15 kHz test tone to check whether a tweeter or your own hearing still responds up there.
Input
sawtooth · 15 kHz · 5 s · 48 kHz / 24-bit · normalize on
Output
sawtooth-15000hz-5s.wav, fundamental only, spurious content 78 dB down
Every harmonic of 15 kHz sits above the 24 kHz Nyquist limit, so a correct band-limited sawtooth contains nothing but the fundamental. A naive generator folds those harmonics back down as false tones (the second lands at 14.1 kHz), which would make you measure the artefact instead of the tone. Export renders the same node graph you auditioned, so the file is the signal, not a re-creation of it.
Generate band-limited test tones, sweeps and noise anywhere from 20 Hz to 20 kHz. Pick a waveform, set the frequency by slider, number or note name, and play. What you export is rendered through the same oscillators you hear, so the file matches the audition exactly. Useful for tuning instruments, testing speakers, calibrating equipment, ear training and tinnitus frequency matching.
Tap the A4 concert pitch preset for a 440 Hz sine, watch the note badge confirm A4 at 0 cents, then play it as a steady reference. Nudge the A4 slider to 432 Hz and the same tone reads as A4 again, since the reference moved with it.
Switch to Sweep, set 20 Hz to 20 kHz over 20 seconds in logarithmic mode, set the export length to 20 seconds and render a 48 kHz 24-bit file. The instantaneous frequency follows f0 times (f1/f0) to the power t/T, which is the sweep shape measurement software expects.
Apply the Meditation (Theta) preset: a 150 Hz carrier in the left ear and 156 Hz in the right, producing a perceived 6 Hz beat. Export renders it as a true stereo file with the two frequencies kept in separate channels.
Run a logarithmic sweep through a system and record it, or step through fixed frequencies to find rattles, nulls and rolloff points.
Play an exact reference pitch with a note name readout and cent offset, at any A4 reference from baroque 415 Hz to modern 440 Hz.
Learn to identify specific frequencies by ear, a useful skill for audio engineers and musicians, with the spectrum display to check yourself.
Render calibration tones, pink noise beds or sweeps as WAV at the sample rate and bit depth your target system needs.
Sweep or step until a tone matches the pitch you hear, then read off the frequency and note name to describe it precisely.
The Tone Generator drives up to eight independent voices through a shared master gain into an AnalyserNode, and the same graph is reused for export. That last detail is the important one: the Export WAV button does not re-synthesize your tone from formulas, it renders the identical node graph on an OfflineAudioContext, so the file you download is the signal you auditioned. Web Audio oscillators are band limited, meaning their harmonics stop at the Nyquist frequency instead of folding back down as false tones. Measured on the shipped build, a 15 kHz sawtooth export carries its fundamental and nothing else above 78 dB down, and the harmonic amplitudes of the square, sawtooth and triangle waves match the ideal Fourier series (1/n for a saw, odd 1/n for a square, odd 1/n squared for a triangle) to within 0.0001.
Each voice has a frequency from 20 Hz to 20 kHz with 0.01 Hz precision, set three ways: a logarithmic slider that gives the musical low end as much travel as the top octave, a numeric field, or scientific pitch notation typed straight in (A4, C#3, Bb5). Alongside the frequency sits a live note readout showing the nearest equal tempered note and its cent offset, and the A4 reference behind that readout is adjustable from 415 to 466 Hz for baroque and alternative tunings. Per voice you also get level, stereo pan and a detune control in cents, plus three noise colours next to the four oscillator shapes: white (flat), pink (falling about 3 dB per octave) and brown (about 6 dB per octave), all generated from a seeded PRNG so an export reproduces exactly what you heard.
Three modes share those voices. Tones is the multi-oscillator view, with one-tap major and minor triad builders that stack three voices from the current root. Sweep glides oscillator one between any two frequencies over 1 to 60 seconds, either linearly (equal time per hertz) or logarithmically (equal time per octave, which is what speaker and room measurements expect); the scheduled ramp tracks the textbook formula to within 0.03 percent. Hearing range steps a sine through a 16 point ladder from 125 Hz to 20 kHz so you can mark where your ears, headphones and speakers stop responding together. Binaural mode is independent of all three: it duplicates every pitched voice, pans the original hard left and a copy pitched 1 to 40 Hz higher hard right, and needs headphones because the beat only forms when each ear receives its own channel.
The analyser above the transport switches between a zero crossing triggered oscilloscope (so a steady tone stands still instead of scrolling) and a logarithmic spectrum display with labelled decade gridlines, both rendered at device pixel ratio so they stay sharp on high density screens. Export options cover length from 1 to 300 seconds, 44.1, 48 or 96 kHz, 16 or 24 bit PCM or 32 bit float, forced stereo, a 0 to 200 ms click free fade, and optional normalisation that lands the peak at exactly -0.5 dBFS. When normalisation is off and the summed voices would exceed full scale, the tool says so rather than silently clipping. The share button encodes the whole patch, every voice plus the binaural, sweep and tuning settings, into the URL.
Yes. Export renders the same node graph on an OfflineAudioContext rather than re-synthesizing the tone from formulas, so the waveform, harmonic content and levels are identical to playback. Only the sample rate can differ, since playback uses your device rate and export uses the rate you pick.
A naively generated square or sawtooth contains harmonics above half the sample rate, which fold back down into the audible range as false tones (aliasing). Band-limited oscillators stop at the Nyquist frequency instead. On this tool a 15 kHz sawtooth export measures its spurious content at 78 dB below the fundamental, so a high frequency test tone is actually a test tone rather than a chord of artefacts.
Yes. Choose a length from 1 to 300 seconds, a sample rate of 44.1, 48 or 96 kHz, and 16 or 24-bit PCM or 32-bit float, then export a WAV. Fades of up to 200 ms avoid clicks at the ends, and optional normalisation puts the true peak at exactly -0.5 dBFS. Everything is generated in your browser.
Up to eight voices, each with its own frequency, waveform, level, pan and detune, so you can layer tones into chords or stack a fundamental with its harmonics. One-tap buttons build major and minor triads from the current root.
Yes. Type scientific pitch notation such as A4, C#3 or Bb5 into the field beside the numeric entry. Each oscillator also shows the nearest note and how many cents sharp or flat it is, calculated against an A4 reference you can move between 415 and 466 Hz.
A linear sweep spends equal time on every hertz, so it races through the low octaves and dwells at the top. A logarithmic sweep spends equal time in every octave, which distributes energy the way hearing and most measurement software expect. The sweep follows the scheduled ramp to within 0.03 percent either way.
Binaural mode plays a slightly different frequency in each ear to create a perceived beat, adjustable from 1 to 40 Hz, with presets for Focus (14 Hz), Relaxation (10 Hz), Meditation (6 Hz) and Deep sleep (2 Hz). Headphones are required: through speakers the two tones mix in the air before reaching your ears and no beat forms.
No. Your headphones, speakers, sound card and volume setting are not calibrated, so the tool cannot know the actual sound pressure level reaching your ears. It shows roughly where your whole playback chain plus your hearing stops responding. For a real assessment see an audiologist.
White noise has equal energy per hertz and sounds bright and hissy. Pink noise falls about 3 dB per octave, giving equal energy per octave, and is the standard choice for speaker level setting because it is closer to how hearing weights frequency. Brown noise falls about 6 dB per octave and sounds like a deep rumble or heavy rain.
Audio is decoded and processed locally with the Web Audio API. Your files are never uploaded to a server.