A clean cut is defined by decoded sample-frame boundaries, so the details that matter are the start and end seconds you type, the decoded duration the browser reports, and the data-size math the tool displays before download. Every WAV you receive from a browser-based trimmer is rebuilt from the Float32 samples that Web Audio decoded into memory, then encoded as uncompressed 16-bit PCM with a fresh 44-byte RIFF header. The original MP3, AAC, FLAC, or Ogg container, along with tags, artwork, cue points, and loudness metadata, is discarded during that rebuild. Knowing which inputs the tool measures (decoded duration, frame count, channel count, sample rate) and which it does not (encoded bitrate, source file size, container format) is what separates a precise, repeatable cut from a cut that surprises you at playback.
Frame boundaries and why seconds are rounded
When you enter "12.500" and "18.250" in the time fields, the browser does not preserve those decimals as audio samples. Web Audio decodes the source into an AudioBuffer, which is a flat array of Float32 frames sampled at the AudioContext rate. Your start and end seconds are multiplied by that sample rate to obtain frame indices, then rounded to the nearest integer frame. The start frame is included and the end frame is excluded, so the selected length equals (end frame − start frame) ÷ sample rate.
This rounding is why the tool displays a cut duration equal to selected frame count ÷ decoded sample rate rather than a timer-based estimate. If your times round to less than one frame apart, the selection is rejected instead of producing an empty file. The production exporter and isolated tests share the same frame-range function, so the displayed cut duration is always the exact frame count divided by the decoded sample rate, never an unrelated timer estimate.
The cut inputs you must get right
Five hard limits must line up before a file is accepted:
- The encoded input file must be no larger than 25 MB.
- Decoded length must be at or under 15 minutes.
- Decoded sample rate must be at or under 192 kHz.
- Decoded channel count must be at or under 8.
- Total decoded samples across all channels must be at or under 30,000,000.
The fifth detail is the one that surprises people working with high-rate multichannel files. A 96 kHz stereo file can hold 30 million channel-samples in roughly 2 minutes 36 seconds, so the per-channel sample budget is reached well before the 15-minute duration cap. The tool does not silently truncate an over-limit file; it rejects the decode with a clear message.
The time inputs themselves are bounded too:
| Input | Rule |
|---|---|
| Start time | Non-negative, strictly less than the decoded duration |
| End time | Strictly greater than start, at most exactly the decoded duration |
| Selection length | Must round to at least one full sample frame |
| Either field | Changing it clears the previous WAV and the prior download URL |
If a value falls outside these rules the field is treated as invalid. Changing either time field clears the previous result, and cutting again replaces and releases the prior download URL.
Cut audio in three steps
- Open Audio Cutter and pick a browser-decodable audio file from the file picker. Wait until the page shows the decoded duration, sample rate, channel count, and frame count — those four numbers describe the working buffer, not the encoded source.
- Enter a non-negative start time and a later end time in seconds. The end time must not exceed the decoded duration shown on the page. Confirm the displayed cut duration matches the segment you intended.
- Select Cut and export WAV. Use the preview player to check that the spoken word or musical transient lands cleanly inside the new boundaries, then download the file. Each download overwrites and releases any prior result URL. If you want a quick method for confirming the boundaries after export, the how to check the result after you cut audio walkthrough covers what to listen for.
What the output WAV contains — and what it drops
The downloaded file is always a brand-new RIFF/WAVE file, not a section of the source. The 44-byte header records:
| Header field | Value written |
|---|---|
| Format code | 1 (PCM) |
| Channels | Decoded channel count, up to 8 |
| Sample rate | Decoded AudioBuffer sample rate |
| Byte rate | Sample rate × channels × 2 |
| Block align | Channels × 2 |
| Bits per sample | 16 |
| Data size | Exact frame count × channels × 2 |
The samples themselves are taken from the AudioBuffer's Float32 channel arrays, clamped to the range −1 through 1, with negative full scale becoming −32768 and positive full scale becoming 32767. Any non-finite value (NaN, Infinity) is written as silence. Channels are interleaved in their original order, so a stereo source produces a stereo result, and a multichannel source keeps its layout up to the eight-channel ceiling. Your playback software must support the resulting channel layout — a stereo-only media player cannot decode a 5.1 file correctly.
What the export drops: the original compressed codec and bitrate, ID3, Vorbis, or FLAC tags, album art, cue points, chapters, loudness metadata, encoder settings, and any other container fields. Treating the WAV as a clean sample dump, rather than a sliced copy of the source, is the right mental model.
File size you can predict before download
PCM16 is uncompressed, so the WAV can be far larger than a short MP3 or AAC source. The data section uses 2 bytes for every selected frame in every channel, plus a 44-byte RIFF header. As one worked example, a stereo file that decodes to 48 kHz and is cut for 30 seconds produces 30 × 48,000 = 1,440,000 frames, then 1,440,000 × 2 channels × 2 bytes = 5,760,000 data bytes, plus 44 bytes of header, for a 5,760,044-byte file — about 5.49 MB. The page reports the output duration and data size before you download so you can confirm the result matches your expectation before committing.
For longer or multichannel selections, the relationship is linear: each additional frame adds channels × 2 bytes, and each additional channel doubles the per-frame cost. A mono source at the same length would produce half the bytes; a 5.1 source would produce three times the bytes of the stereo example above.
Browser-dependent decoding and what to watch for
The file picker recognizes common MP3, WAV, M4A, AAC, Ogg, WebM, and FLAC names and MIME types, but recognition is not a guarantee of decoding. Per the MDN BaseAudioContext.decodeAudioData reference, the actual decode step is performed by the browser's Web Audio implementation, and container and codec support differs across Safari, Chrome, Firefox, operating systems, and browser versions. An unsupported, damaged, mislabeled, encrypted, or incomplete file produces a clear decode error rather than a silent fallback.
Web Audio may also resample the source to the AudioContext sample rate while decoding. The page displays the decoded buffer's sample rate, not the encoded source rate. A 44.1 kHz source opened in a 48 kHz AudioContext is decoded at 48 kHz, and the exported WAV records 48 kHz in its header. If your downstream workflow depends on a specific sample rate, normalize the source to that rate before cutting or resample the exported WAV afterward with a dedicated editor.
Browser decoding necessarily happens before decoded duration and sample count can be inspected, so a highly compressed long file can still require temporary memory during the decode attempt. Nothing is shortened automatically: an over-limit file is rejected. Audio contexts are closed after decoding and during cleanup, and in-flight reads are guarded so an older job cannot overwrite a newer selection.
When a dedicated editor is the better choice
Audio Cutter is built around the single task of trimming one continuous range to sample-frame accuracy and writing a clean PCM16 WAV. It does not offer fades, normalization, gain change, pitch shifting, noise removal, channel remixing, target resampling, or compressed-frame cutting. If your cut needs to land inside a transient (a drum hit, a sung syllable) with sub-frame precision, you need a waveform editor that lets you place the cursor on a zero-crossing. If you need to keep the original codec, tags, or chapter markers, you need an editor that operates on compressed frames or re-tags after export. The dependable habit is to preview the cut WAV, confirm the spoken words or musical transient at both boundaries, keep the original file untouched, and reach for a full editor when the brief calls for more than a clean trim.
Related reading: What Is the Best Way to Cut Audio? A Local Browser Method.