How the page is laid out

It is one page, with no tabs to hunt through. The top half is your list of files. The bottom half is the editor, which works on whichever file you have opened from that list.

Drop in as many recordings as you like. Every one becomes a row. Each row carries a tick box, a play button so you can hear which is which, a drag handle for ordering, a GAP field, an ✎ Edit button.

Ticking a row means exactly one thing: include it when the Join panel stitches the ticked rows, top to bottom, into one continuous track. Converting is not a list job: open the file and use the Save panel, which writes any format, one file at a time.

Getting started

Open the editor at tidyaudio.com/app. Any browser will do. Drop an audio file onto the panel at the top of the editor, or click it to browse. WAV, MP3, M4A, OGG, FLAC and anything else your browser can decode will load, and .amr phone voice memos too, which the tool decodes itself.

Once it loads you get a waveform, a transport bar, and the panels underneath: cut and stitch, clean and level, music overlay, export. The bar at the top shows the length, sample rate, channels, and how large it would be as a 16 bit WAV.

Dropping another file anywhere on the page simply adds it to the list. Nothing is replaced and nothing is at risk, so there is no dialog to answer. Open whichever row you want to work on, and use Close this track to clear the editor when you are done. Your edits stay on the row.

Reading the controls without coming here

Every control in the editor carries a sentence or two explaining what it does and when to reach for it. On a computer, rest the pointer on anything for a moment and it tells you. That covers the transport, the zoom and scroll controls, all the editing buttons, noise, levelling, the music overlay, every export button, and the rows in the list at the top of the page down to the drag handle and the gap field.

A phone has no pointer to rest, and a tap has to run the control rather than describe it, so on a touch screen the tips live behind the ? button in the header. Press it and the editor goes into help mode: every control that has something to say is outlined, and tapping one tells you about it instead of doing it. Press ? again, or the escape key, to go back to normal.

Moving around the waveform

Selecting and placing the playhead

• Click anywhere on the waveform to place the gold playhead. That is where playback and inserts happen.
• Drag across the waveform to select a region. It tints teal, and the readout under the waveform gives you the start, end and length.
• Drag an edge of an existing selection to adjust it. Grab within a few pixels of the teal line.
• Type exact times into the start and end boxes in the Selection panel. Both 1:23.500 and plain seconds like 83.5 are accepted.
• The -0.1 and +0.1 buttons nudge each edge by a tenth of a second, for trimming a breath off the front of a cut.

Zooming and scrolling

Zoom with + and −, or hold ctrl and use the mouse wheel to zoom around the pointer. Sel zooms to fit your selection, Fit zooms back out to the whole file.

Once you are zoomed in, the bar directly under the waveform is your scroller. Drag the thumb to move along the timeline, or click anywhere on the bar to jump there. Holding shift while scrolling the wheel over the waveform does the same thing. A plain wheel scroll is left alone so the page still scrolls normally.

Listening

▶ plays from the playhead, ◼ stops and returns to the start of the selection. ▶ Selection plays only the selected region, and with Loop ticked it repeats, which is the fastest way to audition a cut point. The Speed slider changes playback rate without touching the file.

Cutting a piece out and stitching the rest

This is the main job the tool exists for: get rid of a cough, a long pause, a fumbled sentence, and close the gap.

1. Find the offending bit and drag a selection over it. Zoom in first if it is short.
2. Tick Loop and hit ▶ Selection to confirm you have exactly the right piece, no more and no less.
3. Press ✂ Cut selection and stitch, or just hit delete.
4. The two sides are rejoined with a four millisecond crossfade, so the seam does not click. The playhead lands on the seam. Press play to hear it.
5. If it is wrong, hit the Undo button in the toast that just appeared, or ctrl+z.

The other edit buttons

Undo

Every edit is undoable: cuts, trims, fades, noise cleanup, volume levelling, presence, mixing in music. There are three ways to reach it, all doing the same thing.

• The ↶ Undo button in the transport bar, always visible above the waveform.
• The Undo button inside the toast that pops up after each edit. It stays live for about seven seconds.
• ctrl+z to undo, ctrl+shift+z to redo.

The line next to the undo buttons in the Cut panel tells you how many steps you can go back and what the last one was. History holds fourteen steps, and older ones are dropped early on very long files to keep memory sane. Loading a new file clears the history.

Cleaning out noise

This removes steady background: air conditioning, computer fans, mic hiss, street rumble. It learns what the room sounds like when nobody is talking, then filters that fingerprint out of the whole recording, deciding frequency by frequency and moment by moment how much is noise and how much is you. It will not fix a barking dog or a slamming door, because those are not steady. Silence those by hand instead.

It keeps following the noise as it drifts, so a fan that changes speed partway through is still handled. Reckon on it taking roughly a quarter of the length of the recording to run, so about fifteen seconds for a ten minute file, with a progress bar the whole way.

The way that gives the best result

1. Find a stretch of a second or two where you are not speaking, just the room. The start of the recording usually has one.
2. Select it.
3. Tick Use selection as the noise sample.
4. Set the strength and press ✧ Clean out noise. A progress bar runs while it works, roughly a quarter of the audio length on a normal machine.

Without a selection it finds the quietest stretch on its own, which is usually fine. If the recording never has a quiet moment it will tell you so and ask you to pick one.

Choosing the strength

On a normal recording, where the background sits well below the voice, all three settings leave the voice itself within about a decibel of where it was. The setting decides how hard the background is pushed down, not how much the voice is mangled. A recording where the noise is nearly as loud as the speech is a different matter, and no setting will come out of that cleanly.

Evening out the volume

Three buttons, three different jobs. They are easy to confuse, so here is the difference.

≡ Equalize volume

This is the one you usually want. It finds each spoken passage, works out how loud that passage is, and gives the whole passage one steady gain that brings it to the target. The Target slider sets that loudness. Around -18 dB suits spoken word, quieter if you want more dynamics, louder for something that will be listened to in a car.

Two things follow from working passage by passage rather than continuously. The gain only ever moves during the gaps between phrases, where there is nothing to hear, so you never get the breathing or pumping that a continuous leveller produces. And each passage genuinely reaches the target rather than being dragged part of the way, so a recording that wandered by twenty decibels comes out flat to within about one.

Room tone in the gaps is never lifted on its own account, and the target is absolute. Run it twice at -18 and the second run does nothing. Run it at -18, change your mind and run it at -24, and you land on -24, not on some compound of the two. Anything that would have clipped is softly limited, and only if it would actually have clipped.

↑ Normalize peak

Multiplies the whole file by one number so the single loudest moment lands just under maximum. It changes nothing about the balance between loud and quiet parts. Use it as a last step before export, or on a recording that is already even but simply too quiet.

+ Add presence

A high shelf that lifts everything above roughly 6 kHz by 2 dB and leaves everything below it alone. On a voice that means consonants get crisper and the recording sounds closer, more like someone in the room and less like someone behind a blanket. It is the same lift the old practice builder applied automatically to every build. It is a button now so you can hear it and undo it, rather than it happening silently to everything.

Reach for it when a recording sounds dull or boxy, or came off a mic that rolls away the top end, or was taken on a laptop or a phone. It also helps after a heavy noise pass, since cleaning tends to take some air with it.

Leave it alone when the voice is already clear, and especially when it is sibilant. A lift at 6 kHz sharpens ess and tee sounds along with everything else, so on a hissy or sibilant take it makes matters worse.

Laying music under the track

The Music overlay panel holds a second stream alongside your recording. It stays separate, and adjustable, until you press mix, so you can move it, change the volume and hear the result before committing to anything.

1. Press Load music or second track and pick your file. It appears on the waveform in purple, drawn at whatever volume you have set, with a purple line marking where it begins.
2. Decide where it starts: type a time, or press At playhead or At start.
3. Set the Volume. Music sitting under a voice usually wants somewhere between -20 and -10 dB. The default is -14.
4. If the music is shorter than the recording, tick Loop it to fill the track. Each loop point gets a tiny fade so there is no click. If it is longer, it gets cut off at the end unless you tick Let it run past the end, which extends the track instead.
5. Set the Fades, in seconds, at the start and end of the music.
6. Leave Duck the music under the voice ticked. The music drops by the amount you set whenever you are speaking and comes back up in the gaps, quickly down and slowly back, the way a radio bed behaves. Untick it if the overlay is not music but a second voice you want at a constant level.
7. Press ▶ Preview the mix and listen. Nothing has changed yet. Adjust and preview again as many times as you like.
8. When it sounds right, press ⊕ Mix the overlay into the track. Now it is part of the recording, and like any edit it can be undone.

Splitting one file into parts

This is what the old splitter tool did, now in the Export panel.

1. Click on the waveform to put the playhead exactly where the split should happen. Play around it to be sure.
2. Choose WAV or MP3 at the top of the Export panel.
3. Press ↓ Part 1 (start to playhead), then ↓ Part 2 (playhead to end). Two files, saved separately.

To pull a single chunk out of the middle instead, select it and press ↓ Selection only. For more than two parts, export each range with the selection button, or split once and reload each half.

The list at the top of the page: joining

Everything that involves more than one file happens in the list at the top of the page. Drop in as many recordings as you like, tick the rows you want, and join them into a single track.

Each row carries what you need to work with it:

• A tick box: include this row when joining. Ticks do nothing else.
• A play button, so you can hear which recording you are looking at.
• A drag handle for reordering.
• A GAP field: seconds of silence after that row when joining.

Joining recordings end to end

1. Drop the files in and drag them into the order you want.
2. Set the GAP on each row for a short pause after it. For a longer stretch of quiet, set the minutes at the top and press Add silence: it drops in as its own row that you can drag between any two files, and its length stays editable in place afterwards.
3. Untick anything you want left out. Nothing is deleted by unticking.
4. Leave Match levels ticked if the recordings were made at different volumes. Tick Crossfade joins if you want pieces to blend rather than butt together, which applies wherever the gap is zero.
5. Press Join into one track.
6. Download it, or open the new row to keep working on the joined track.

One way in from the editor: Add the selection to the list above in the Export panel lifts the selected region out as its own row, so a piece of one recording can be joined between two others without exporting anything in between.

Changing a file from one format to another

There is no separate converter to go and find. For the file you are working on, converting is what the Export and convert panel does, because by then the file is open and you can hear it. Drop the file in as you would for any other job, play it to be sure it is the right one, then set what you want out:

• Format, WAV or MP3. Choosing MP3 reveals a quality setting.
• Sample rate, left on same as the track unless you have a reason. Dropping a voice recording to 22050 or 16000 makes it much smaller, and the browser does the resampling rather than anything homemade here.
• Channels, folding stereo to mono where the recording is one voice on both channels. That halves the size for nothing.

The line beside the channels setting tells you roughly how big the result will be before you commit to it. Press Export edited file and you have your conversion. Nothing about the loaded track changes, so you can export the same recording three different ways without reloading it.

Reading a file in depends on your browser, which in practice covers WAV, MP3, M4A, OGG, FLAC and more, plus .amr phone voice memos, which the tool decodes itself. Writing is WAV or MP3 only, since those are the two encoders that fit in a single page with nothing to install. An AMR is 8000 Hz mono by nature, so expect a telephone sound, and converting it cannot add back what the phone never recorded.

Several files?

Convert them one after another: open a file, save it in the format you want, open the next. Each file keeps its own name and its own settings, and you hear what you are converting before you commit to it.

Exporting

Where files end up

Anything you save puts up a proper Save As dialog, so you choose the folder and the name. For an export the dialog appears before the encoding starts rather than after, because a browser only lets a page open one while the click that asked for it is still fresh. Cancel it and nothing is encoded and nothing is written.

Two cases fall back to an ordinary download, straight into whatever folder your browser is set to use: saving several converted files at once, since a browser allows only one dialog per click, and any browser that does not offer the file picker at all. Names are built from the file name plus a suffix: _edited, _selection, _part1, _part2.

Recipes

Tidying up a raw recording

1. Load it. Select the dead air at the front, Trim to selection around the real content.
2. Work through the fluffs: select, loop-audition, cut and stitch.
3. Select two seconds of room tone, tick the noise sample box, clean at Medium.
4. ≡ Equalize volume at -18 dB.
5. Fade in over the first second, fade out over the last two.
6. Export as MP3.

A practice session with music under it

1. list at the top of the page: drop in your sections, order them, set gaps, add a silence row where you want quiet practice time, tick them all and join.
2. Open the new Joined track row that appears at the bottom of the list.
3. Load your music in the overlay panel, start at 0, loop on, ducking on, volume around -16 dB.
4. Preview, adjust, mix.
5. Normalize, then export.

Cutting one long recording into episodes

1. Load it and clean and level the whole thing first, so every piece matches.
2. Place the playhead at the first break, export Part 1.
3. Select the next range and use ↓ Selection only for each following piece.

Keyboard shortcuts

These work in the editor when you are not typing in a box.

If something goes wrong

The file will not load

The message will say it could not decode. Browsers cannot open every format, and some WMA and unusual M4A files are refused. Convert it to WAV or MP3 elsewhere and try again.

Nothing plays

Browsers block audio until you interact with the page, so click something first. Also check the Speed slider has not been left somewhere odd, and that a selection is not the reason you are only hearing a short burst.

The cut left a click

Undo, zoom in further and place the cut in a genuinely silent moment rather than mid-word. The four millisecond crossfade handles most seams, but cutting across a vowel will still be audible.

Everything got slow

Long files are heavy, and every undo step holds a copy. Export what you have, reload the page, and load the exported file back in to start with a clean history. Working in pieces and joining them in the list at the top of the page at the end is usually faster than editing one enormous file.

The overlay music sounds too loud in the pauses

That is ducking working correctly, the music comes back up when you stop talking. If the jump is too obvious, make the duck amount smaller, like -6 dB instead of -12, so the difference between ducked and open is less dramatic. Or lower the overlay volume overall.

I closed the page by mistake

Nothing is saved between sessions, so unexported work is gone. Get in the habit of exporting a WAV once the heavy edits are done, before starting the fiddly bits.