The workshop · Trim a recording
How it works : Trim a recording
Your phone can already record. What it cannot do is cut the start and the end without sending the recording to somebody: that is exactly what this page does, and the file it returns carries no metadata at all.
A voice recording is one of the most intimate things we handle: a meeting, a lecture, a memo, an interview. The sites that cut them up receive them whole, and nobody knows what they do with them. The browser, meanwhile, can decode a sound and rewrite a piece of it without a single byte leaving.
The tool does something most trimmers do not: it SHOWS what your file carries before cleaning it. An audio file often carries metadata chunks, holding the name of a program, a title, sometimes an author's name. They are listed on screen, then they vanish from the file produced, and the tool reads its own result back to check it.
One thing has to be said straight away. This is not a dictaphone: it does not record from a microphone, and the reason is written in the limits. And there is no waveform to cut with the mouse: the start and the end are set in seconds, because drawing a signal would be a graphical component this site's design system does not have.
How it works
- Drop your recording in. The tool decodes it and shows its length, its sample rate and its metadata.
- Set the start and the end in seconds. The end comes pre-filled with the total length.
- The file you get is a WAV in 16-bit PCM, with no metadata at all, and the tool reads its own output back to tell you what it holds.
Formats accepted
Which formats can be decoded depends on the browser, and no first-hand source documents that precisely: the tool TRIES to decode and tells you plainly if it could not. The result is always a WAV.
WAV, uncompressed
It is the only format whose header the tool can read BEFORE decoding, which lets it keep the original sample rate exactly. It is also the only one whose metadata it can list.
MP3
Decoded by every browser. The sample rate cannot be read before decoding, so it will be brought to whichever rate you choose, and the tool says so.
M4A and AAC
The format an iPhone's recorder produces. Decoding depends on the browser AND the system: if yours cannot, the tool says so rather than handing back an empty file.
WebM and Opus
The format a Chromium browser's recorder produces. Decoded everywhere except by the oldest browsers.
Its limits
This is not a dictaphone: it does not record from a microphone
Two reasons, and the second is the real one. First, the anatomy of a tool page on this site offers ONE input per tool: a file, a text, or nothing. A microphone is a fourth kind of input, which would need a fresh variant of the trust banner, of the message for browsers without JavaScript and of all three test locks. Second, and above all: the machine that builds this site has no microphone, so no lock could exercise that input, and the house rule is explicit, a path you cannot exercise does not get written. Your phone and your computer can already record; what they cannot do is trim without uploading.
There is no waveform to cut with the mouse
Drawing a signal would be a graphical component this site's design system does not have, and the design is frozen: what is missing gets noted and requested, it does not get invented in the middle of a tool. So the start and the end are set in seconds, to the hundredth. It is less pleasant, and it is said here rather than discovered.
The result is always a WAV, therefore heavier than the input
A minute of sound as WAV weighs about ten megabytes, against one for an MP3. That is the price of having no dependency: the browser can now encode audio, but its encoder returns encoded chunks, never an assembled file, and lossless encoding is almost nowhere. Writing a WAV container by hand is therefore the only route that works everywhere, and a WAV reopens in every editing program.
A file that is not a WAV is RESAMPLED, and the tool says so
The specification of the browser's audio engine requires decoding to bring the sound to the rate of the computing context. For a WAV, the tool reads the rate in the header BEFORE decoding and creates the context at that rate: nothing is resampled. For other containers the rate cannot be read without decoding, so the sound is brought to the rate you choose, and the difference is shown on screen.
120 MB per file
Decoded sound lives in memory as floating-point numbers, which weighs several times the original file. Past this limit a phone crashes the tab instead of returning a result, and the tool refuses cleanly, stating the limit.
Possible sample rates run from 8,000 to 96,000 Hz
That is the range the specification guarantees in every browser. Outside it, the browser refuses to create the computing context. It is amply enough: compact disc quality is 44,100 Hz, cinema and video 48,000 Hz.
Questions about this tool
Is my recording sent anywhere?
No, and this is the tool where it matters most, because a voice recording identifies the person speaking. Open the inspector with the F12 key, go to the “Network” tab, clear the list, then drop your file in and trim: nothing leaves. Cut your connection and do it again, the tool still works.
Why is the file I get heavier than the original?
Because a WAV is not compressed: it holds every sample in the clear. A minute weighs about ten megabytes, against one for an MP3. We cannot do otherwise without embedding an encoder, and this site only embeds a library on a reasoned decision. The browser has recently learned to encode audio, but its encoder returns encoded chunks you must assemble into a container yourself, and lossless encoding is almost nowhere. Meanwhile a WAV opens anywhere, edits without loss, and hides nothing.
Why not record straight from the microphone?
Because this page would not be as honest. Every tool on this site has ONE input, and its trust banner speaks about exactly that input: your file, your text, or what it builds. A microphone would be a fourth kind of input, with its own promise to write. Above all, the machine that checks this site before every release has no microphone: we could therefore not PROVE that a recording does not leave, and we do not publish a promise we cannot exercise. Your device already has a recorder that works; it was trimming without uploading that was missing.
What exactly does the metadata cleaning do?
It removes nothing: it writes nothing. The file produced is composed from scratch, with two chunks only, the one saying how the sound is encoded and the one carrying the sound. Everything else, program name, title, comment, player tags, is simply not copied. The tool first shows you what your file carried, then reads its own output back chunk by chunk to check nothing is left.
Why did my sample rate change?
Because your file was not a WAV. The specification of the browser's audio engine requires decoding to bring the sound to the rate of the computing context, and that rate has to be chosen BEFORE decoding. For a WAV we read the rate in the header and create the context at that rate: nothing changes. For an MP3 or an M4A the rate cannot be read without decoding, so we use the one you chose, and we show the difference. Resampling does not damage a speech recording much, but it is not neutral, and hiding it would be dishonest.
Can I cut a piece out of the MIDDLE of a recording?
Not in one go: the tool keeps one continuous interval, from the chosen start to the chosen end. To remove a passage in the middle, export the two pieces surrounding it, then join them in an editing program. Joining two pieces would mean handling transitions, levels and ordering, which is a different tool with a page of its own.
Your phone can already record. What it cannot do is cut the start and the end without sending the recording to somebody: that is exactly what this page does, and the file it returns carries no metadata at all.
VerifyThe proof, step by stepThe network inspector check, explained one gesture at a time, with screenshots.
The cornerThe everyday cornerConvert units and see where every factor comes from, work out dates without being tripped up by a clock change, trim a recording, zip a folder. The tools you look for once a month and never find again.