The sound a product makes when something works, fails or needs attention. Pick the situation, nudge it until it fits, and download it — synthesised here, so there is no sample licence attached to it.
A sound effect is a message before it is a sound. Rising means it worked, falling means it did not, two notes a semitone apart mean something is wrong, and a steady repeated beep means look at this now. Those associations are shared by everyone who has ever used a computer, which is why the presets here are organised by situation rather than by waveform: the useful question is not "what does a square wave sound like" but "what does a partial success sound like".
Everything is synthesised from the numbers on the sliders — an oscillator on a phase accumulator, an exponential envelope, a one-pole filter — evaluated sample by sample in the page. Nothing is recorded and no sample library is involved, so what comes out carries no licence and needs no attribution, and the waveform drawn above the controls is the exact array of samples the downloaded file is built from.
Yes. Nothing here is a recording and nothing is drawn from a sample library, so there is no licence attached to what comes out. The waveform is computed from the numbers on the sliders by code running in your browser, which makes the result yours to ship, sell or modify without attribution.
Shorter than feels right when you are auditioning it. A click or a confirm wants 60–150 ms; a notification 200–400 ms; a success 400–600 ms. Anything past about 700 ms stops being feedback and starts being an event the user has to wait through, and the ones people mute first are always the long ones. The length is shown next to the play button, and Speed is what changes it.
Because a sound effect is a message, and the vocabulary is already established. Rising means good, falling means bad, a semitone apart means something is wrong, and a steady repeated beep means look at this now. Those are learned associations shared by everyone who has used a computer, and a product that inverts them is not being original — it is being misunderstood. The controls are deliberately weighted towards timbre, which you can make your own, rather than towards direction, which you probably should not.
Tone is a lowpass filter: turning it down removes the high harmonics, so a bright chime becomes a muffled thud. Decay shapes each note — at 0 it holds flat like an organ, at 100 it is plucked and gone. Retro reduces the bit depth from 16 towards 3, which is the quantisation noise that made 8-bit consoles sound the way they did. Noise blends white noise into the tone, and at 100 the pitch stops mattering at all, which is how an impact is made. Detune doubles the oscillator slightly out of tune for thickness, and Echo is a feedback delay tied to the note length.
It bends into the note instead of starting it again. A power-up is one note sliding upwards over an octave, not eight separate notes, and the difference is entirely this switch. Technically it interpolates exponentially in frequency — which is linear in pitch, the only interpolation that sounds like a slide rather than a swoop — and it suppresses the attack, because a note that slides and then re-articulates is heard as two notes.
WAV for anything going into a game engine, an editor or an app bundle: it is uncompressed, it is exact, and every tool opens it. These sounds are well under a second, so the size difference that matters for music is irrelevant here — a half-second WAV is about 44 kB. Choose MP3 or OPUS only if the sound is being served over a network to a browser.
No. There is no microphone involved and no server to send anything to. The sound is arithmetic — an oscillator, an envelope and a filter evaluated sample by sample in this tab — and the file you download is built from that same array of numbers you can see drawn above the controls.