AudioSnap
Record from your microphone or open an audio file, cut it on the waveform, level it, and export lossless WAV or a compressed clip. Everything runs inside this tab.
Record from your microphone
The mic only opens while you are recording, and the audio never leaves this tab.
Microphone settings
Turn the three processors off for music or room tone; leave them on for speech in a noisy place.
How it works
Record or open a file
Pick your microphone and decide whether the browser's noise suppression helps or hurts, then hit record. Or drop in an MP3, WAV, M4A, OGG, FLAC or the audio of a video — nothing is decoded until you ask for it.
Cut on the waveform
Drag the in and out markers, drag across the wave to select, roll the wheel to zoom in on a single word. The playhead, the selection and the zoom are all independent.
Level it and check it
Normalise the peak, add fades, change the speed, fold it to mono. Every setting is a number, not a new copy of the audio, so undo always takes you back and the source never degrades.
Export what you need
16-bit dithered WAV for editing elsewhere, 24-bit or 32-bit float to keep the headroom, or a compressed Opus clip when the file has to be small. The estimated size is shown before you commit.
Sample-accurate trimming
The cut points are seconds in the decoded buffer, not positions in a media element, so the export starts and ends exactly where the markers sit — no quarter-second overshoot.
Real microphone control
Choose the input device, switch echo cancellation, noise suppression and automatic gain off one by one, pick mono or stereo, and set the recording bitrate before you start.
Non-destructive editing
Gain, normalisation, fades, speed and channel changes are stored as a small description of the edit and applied once at export. Undo and redo cost nothing in memory.
Measured, not guessed
Peak and RMS of the selection in dBFS, the gain the render will actually apply, the resulting peak, and a warning before the output would clip.
Exports worth keeping
16-bit WAV is written with TPDF dither so quiet fades stay clean, 24-bit and 32-bit float keep the headroom, and the sample rate is yours to choose.
Nothing is uploaded
Decoding, drawing, rendering and encoding all happen in this tab with the browser's own audio engine. No server, no account, no third-party library fetched from a CDN.
Record, Cut and Export Audio Without Uploading Anything
AudioSnap is a microphone recorder and a waveform trimmer in one page. It decodes your file with the browser's own audio engine, draws a real min/max envelope you can zoom into, and renders the finished cut offline — so the audio never travels anywhere.
One decoded buffer, rendered on demand
Your file is decoded once into raw samples and then left alone. Trimming, gain, fades, speed and channel changes are kept as numbers and applied in a single offline render when you export, which is why undo is instant and repeated edits never stack up quality loss.
A recorder for the takes, an editor for what comes after
Capture a voice memo and cut the dead air off both ends. Pull a clean twenty seconds out of an hour-long interview. Level a voiceover before it goes into a video edit. Fold a stereo recording down to mono for a phone system, or drop a 48 kHz take to 16 kHz for a speech model. The waveform zooms down to individual words, so the cut lands where you meant it to.
Microphone access is requested when you press record and released when you press stop. The recording, the decoded samples and every export live in this tab's memory until you close it. There is no upload endpoint, no analytics on your audio, and no model or codec fetched from a third-party CDN.
Frequently Asked Questions
QCan I trim an audio file I already have, or only recordings?+
Both. The 'Open a file' tab takes MP3, WAV, M4A/AAC, OGG, Opus, FLAC and WebM, plus the audio track of a video file. Dropping a file does not decode it: it waits until you press the load button, so a long recording never freezes the page behind your back.
QIs there a recording length limit?+
There is no limit built into the tool, only your device's memory: the audio is held as raw samples, which is roughly 10 MB per minute of stereo at 48 kHz. Long sessions are fine on a desktop; on an older phone, keep it to a few minutes at a time.
QIs any of this uploaded?+
No. Recording, decoding, drawing, rendering and encoding all use APIs built into your browser. Nothing is sent to a server, and no library, model or codec is downloaded from a CDN while you work.
QWhich export format should I pick?+
16-bit WAV is the safe default and what other editors expect. 24-bit and 32-bit float keep the extra headroom if you plan to process the file again. The compressed Opus option makes a much smaller file, but browsers have no offline audio encoder, so it re-encodes in real time — a two-minute clip takes about two minutes.
QWhat does the 'hold: original' button do?+
It plays the untouched source while you hold it — no trim, no gain, no fades — and returns to your edited version when you let go. Holding Alt does the same thing. It is the only reliable way to tell whether the normalisation actually helped.
QWhat is the difference between gain and normalise?+
Gain is a fixed amount you choose, in decibels. Normalise measures the loudest peak in your selection and works out the gain needed to bring it to the ceiling you set, so two clips recorded at different distances end up at the same level. If you enable both, the manual gain is applied on top of the normalised level.
QWhy does the browser ask for microphone permission?+
Capturing audio always needs your explicit consent, and the browser handles that prompt itself. AudioSnap only asks when you press record, and the microphone indicator goes out as soon as you stop.