Trim MP3, WAV, M4A and more on a visual waveform — make ringtones, cut intros, clip podcasts. Add fade in/out and export MP3 or WAV, all in your browser.
Drop an audio file here, or browse
MP3, WAV, M4A, OGG… · stays on your device
🔒 Decoding, trimming and encoding all run on your device — your music is never uploaded.
Clip the chorus, add a quick fade, export — a custom ringtone without installing an app or uploading your music.
Drag handles on the real waveform so you cut exactly on the beat — and preview the selection before exporting.
Cut podcast quotes, voice-memo highlights, soundboard clips or practice loops in seconds.
Your file is handed to the browser's own audio decoder and turned into raw samples — an uncompressed stream of floating-point numbers, one per sample per channel. Everything after that point works on those numbers rather than on the compressed file, which is why the waveform appears at once, why dragging a selection is instant, and why zooming can go all the way down to individual samples.
The waveform drawing changes technique depending on how far in you are. Zoomed out, each pixel column represents thousands of samples, so the tool finds the highest and lowest value in that range and draws the bar between them — the familiar dense waveform shape, which shows you the true peaks rather than a smoothed average. Zoom past the point where there are more pixels than samples and it switches to drawing the actual waveform as a continuous line, and further still it marks each individual sample as a dot. The maximum zoom fits 50 milliseconds across the whole canvas.
Preview plays from the original decoded audio with the fades applied live, so what you hear is what the export will contain. Clicking the waveform places the playhead; playback resumes from there rather than from the start of the selection, which makes checking a specific moment much quicker than restarting each time.
Every gesture the waveform accepts is decided by where you press. Press on empty waveform and drag, and you draw a new selection. Press on one of the two edges — anywhere within about a centimetre of it — and you resize that edge. Press inside the selection and drag, and the whole region slides while keeping its length, which matters more than it sounds: most re-edits are not "make this longer", they are "same length, slightly later", and rebuilding a carefully-sized region from scratch to move it half a second loses the size you had. Shift-click extends whichever edge is nearer, the way a text selection behaves. Two fingers pinch to zoom and slide to pan.
A plain click places the playhead, including a click inside the selection, and playback always resumes from the playhead rather than from the start of the region — so checking one specific moment does not mean listening to everything before it again. The playhead's flag can be dragged along the ruler to scrub. I and O set the start and end to wherever the playhead is, which is the fastest way to work: listen, stop at the right instant, press a key. Every change to the selection is undoable with Ctrl+Z.
Snap moves each cut point to the nearest place the waveform crosses zero, searching about ten milliseconds either side. This is worth understanding rather than just leaving on. Audio is a wave swinging above and below a centre line; cutting at a moment when the wave is high leaves the clip starting at that value instead of at silence, and a sudden jump from nothing to a large value is a click — a real, audible tick at the edit. Landing on a zero crossing removes the jump at its source. The search window is deliberately small so the edit stays where you put it, and if no crossing is nearby the original point is kept rather than dragging the cut somewhere surprising.
The two modes differ more than they look. Keep selection renders exactly the region you marked — start to end, with fades applied at the boundaries. Remove selection is the harder case: it takes everything before the selection and everything after it and has to splice them together, which is where an audible click comes from if it is done naively, because the waveform jumps instantaneously from one amplitude to a completely different one.
To avoid that, the two halves are joined with a 15-millisecond crossfade — short enough that nobody perceives it as a fade, long enough for the discontinuity to disappear. If you are cutting a cough out of a podcast or an unwanted bar out of a loop, this is what stops the edit announcing itself. Turn on Loop to audition the join repeatedly, which is the fastest way to tell whether an edit reads as deliberate.
Normalise is worth understanding before you use it. It finds the loudest single peak in the finished clip and scales everything so that peak sits at −1 dBFS, just under the maximum. It never turns anything down, and it will not amplify a clip by more than twenty times. Note that this is peak normalisation, not loudness normalisation — it does not measure perceived volume, so a quiet recording with one loud door slam in it will barely move, because that slam is already near the ceiling. For a quiet voice recording with no transients, it works exactly as you would hope.
Memory is the real constraint, and it is larger than the file. Decoded audio takes four bytes per sample per channel regardless of how compressed the source was. A five-minute stereo track at 44.1 kHz occupies about 106 MB once decoded; a one-hour podcast is around 1.27 GB. A 3 MB MP3 becoming 100 MB of memory is normal and expected. Long recordings will be slow or will fail outright on a phone — for those, cut in stages or use a desktop editor.
MP3 export re-encodes. Cutting an MP3 here means decoding it and compressing it again, which is a second lossy pass — the audio is very slightly worse than the source even at the same bitrate. Specialist tools can trim an MP3 on its frame boundaries without re-encoding, at the cost of only being able to cut at roughly 26-millisecond intervals and not being able to apply fades at all. This tool takes the other trade: sample-accurate edits, fades and crossfades, with one generation of loss. Export WAV if the clip is going into another editor, since that avoids the second pass entirely.
Format support is your browser's. MP3 and WAV work everywhere. M4A, AAC, OGG and FLAC depend on the browser and platform, so a file that opens in one may be refused in another. DRM-protected purchases cannot be decoded at all. Video files are not the target here — pull the audio out with video to MP3 first.
One region, one pass. There is a single selection, so removing three separate sections means three rounds of cut-and-re-import. There is no multi-track mixing, no equalisation and no noise reduction — for cleaning up a recording rather than trimming it, the normaliser and silence remover are the neighbouring tools. Fades are linear and capped at 1.5 seconds, or half the clip length for anything shorter. Surround-sound sources are reduced to two channels on export.
Zoom before you place a boundary. At full view a three-minute track spans perhaps 800 pixels, so a single pixel covers around a fifth of a second — far coarser than the ear. Set the selection roughly by dragging, then use Zoom to selection and adjust the handles at high magnification, where the waveform's actual shape shows you where a word or a beat begins. The handles also respond to the arrow keys, and the step size scales with the zoom level, so a nudge is proportionally finer the further in you are. For known timings, type them directly as 1:23.5 or as plain seconds.
Cut on silence wherever you can — the gap between words, the space before a downbeat. An edit placed in a gap is almost impossible to hear; one placed mid-syllable never sounds right no matter how good the crossfade. When there is no gap, a short fade in or out covers the transition.
| Export | Roughly per minute | Use for |
|---|---|---|
| MP3 128 kbps | ~0.9 MB | Speech, voice memos, anything going into a messaging app. |
| MP3 192 kbps | ~1.4 MB | The default. Fine for music in most listening situations. |
| MP3 320 kbps | ~2.3 MB | Music you care about, or a clip that may be re-encoded again later. |
| WAV (16-bit) | ~10 MB | Further editing, or any workflow where a second lossy pass is unacceptable. |
For a ringtone, keep it around 30 seconds, use a fade in so it does not start abruptly, and export MP3. If you are assembling several clips rather than trimming one, the audio merger joins them, and the rest of the set is on the audio & video page.
Drag across the waveform to mark a region, then drag either edge to adjust it. Dragging inside the region slides the whole selection without changing its length — usually what you want when a clip is the right length but starts half a second too early. Press I and O to set the start and end to the playhead, or type exact times into the Start/End boxes (like 1:23.5). Ctrl+Z undoes any selection change. Zoom in first — at full view a single pixel can cover a fifth of a second, which is far coarser than you can hear.
It moves each cut point to the nearest place the waveform crosses zero, looking about ten milliseconds either side. Cutting where the wave is part-way through a swing leaves a sudden step in the audio, and a step is heard as a click at the edit. Snapping to a zero crossing removes it. The search stays small so the cut lands where you put it, and if there is no crossing nearby your exact point is kept.
MP3 for ringtones and sharing (small, plays everywhere). WAV when you'll keep editing the clip or want lossless quality. There is a real reason to prefer WAV for onward editing: exporting MP3 re-encodes the audio, which is a second lossy pass on top of whatever compression the source already had.
Fade in ramps the volume up over the first 1.5 seconds of the clip; fade out ramps it down over the last 1.5 seconds — no clicks, no abrupt cuts. On a clip shorter than three seconds each fade is shortened to half its length so the two never overlap.
There is no hard cap, but the whole file is decoded into memory first, so duration matters far more than file size. Decoded audio needs four bytes per sample per channel — roughly 106 MB for a five-minute stereo track and around 1.27 GB for an hour, no matter how small the compressed file was. Multi-hour recordings will be slow on a laptop and are likely to fail on a phone.
It should not. Splicing two pieces of audio together creates a sudden jump in the waveform, which is what produces a click, so Remove selection joins the two halves with a 15-millisecond crossfade — too short to hear as a fade, long enough to smooth the discontinuity. Turn on Loop to audition the join before exporting.
It scales the whole clip so its single loudest peak sits at −1 dBFS, just below the maximum. It only ever raises the level, never lowers it. Because it measures peaks rather than perceived loudness, a quiet recording containing one sharp noise will barely change — that noise is already close to the ceiling. On an evenly quiet voice recording it works well.
Trim the section you want — usually the chorus — to around 30 seconds, add a short fade in so it does not start abruptly, and export MP3. That file works as a ringtone on Android and in most phone software. iPhone is the exception: iOS wants an .m4r, so the MP3 has to be converted and added through your computer's phone software.