Free tool · MP3 copied, not re-encoded · Runs in your browser
Cut an audio file
Short answer
Choose a start and an end, and the tool keeps just that stretch. An MP3 is cut by copying its frames, so it is not encoded again: our 128 kbps test recording came out at 129 kbps, 0.16 MB for 10 s, with samples identical to the source. Any other format is saved as MP3. Nothing is uploaded.
- Selected length
- 00:00:30
An MP3 is cut by copying its frames, so the sound is not re-encoded and the cut lands on the nearest 26 ms frame. Other formats are saved as MP3. Nothing is uploaded. Maximum 100 MB; the first run loads a ~30 MB engine, once per visit.
Key takeaways
- An MP3 is not re-encoded. Cutting 5 to 15 seconds of our MP3 gave 160,705 bytes at 129 kbps, with decoded samples identical to the source's, so there is no second loss.
- The cut lands on an MP3 frame, which is 1152 samples, 26.1 ms at 44.1 kHz. Our 5 second start landed 15.5 ms from the exact point and the output was 10.005 seconds for a 10 second request.
- Other formats become MP3 at LAME quality 2. Our lossless WAV excerpt came out at 101 kbps mono, with a signal-to-distortion ratio of 28.9 dB against the original (higher is better).
- It is fast: 0.03 s to copy 10 s of MP3, 0.12 s to encode the same length from WAV.
- Start and end accept seconds (90), minutes and seconds (1:30) and hours, minutes and seconds (00:01:30.5). The end is cut back to the length of the file. Files up to 100 MB are accepted.
What the tool does to your file
It reads the length of the file in your browser, checks your start and end against it, and runs one ffmpeg command. For an MP3 the command copies the compressed frames between your two times. For every other format it decodes the sound and saves the stretch as an MP3 at LAME quality 2, a variable-bitrate quality setting.
| Input | Command part | What it means |
|---|---|---|
| MP3 | -ss start -to end -c copy | Frames copied, no decoding, no new loss |
| WAV, M4A, OGG and other formats | -ss start -to end, libmp3lame -q:a 2 | Decoded and saved as MP3 |
| Cut points | Written with three decimals | Millisecond resolution before the frame limit |
| Limits | Up to 100 MB; end cut back to the file length | A start past the end is rejected |
Measured: four cuts
Measured on 30 seconds of a LibriVox reading of The Heart of a Mystery (public domain), as published (128 kbps mono MP3) and as a lossless WAV decode of the same recording. The tool's own ffmpeg arguments ran on the same single-thread WebAssembly engine your browser downloads; times are from a headless run and your device will differ.
The first two rows cut the middle of the recording from start and end timestamps; the last two keep the first 10 seconds, the shape of cut the older version of this tool offered.
| Source and cut | Output length | Output size | Bitrate | Against the source | Time |
|---|---|---|---|---|---|
| MP3, 5 s to 15 s | 10.005 s | 160,705 bytes | 129 kbps (copied) | Identical samples, starts 15.5 ms off | 0.03 s |
| WAV, 5 s to 15 s | 10.031 s | 126,639 bytes | 101 kbps (encoded) | SDR 28.9 dB | 0.12 s |
| MP3, first 10 s | 10.031 s | 161,123 bytes | 128 kbps (copied) | Identical samples | 0.01 s |
| WAV, first 10 s | 10.031 s | 129,385 bytes | 103 kbps (encoded) | SDR 29.5 dB | 0.10 s |
SDR is the signal-to-distortion ratio of the cut against the same stretch of the original; an MP3 copy has no distortion to measure. The MP3 source was 481,279 bytes for 30 seconds.
Why an MP3 is copied instead of encoded again
Every time a sound is saved as an MP3 some of it is thrown away. Encoding a cut again would stack a second loss on the first, and on a speech recording at 128 kbps it would also change the bitrate. Copying the frames avoids both. The price is the frame: ffprobe reports that every packet of our MP3 holds 1152 samples, 26.122 ms, so a copy can only start and end on those boundaries.
That limit is small. Our request for exactly 5 seconds began 15.5 ms away, inside one frame, and the result was 10.005 seconds for a 10 second request. A cut on a word boundary will not be audible.
When the output is a new MP3
A WAV, M4A or OGG file has to become something. The tool saves MP3 at LAME quality 2, which for our mono speech meant about 101 kbps, and the cut scored 28.9 dB SDR against the lossless original. That is a lossy encode, so keep the original file, and cut from it once.
If you need a lossless result, cut with an editor that writes WAV or FLAC. This tool writes MP3.
Free tools:Normalize audio loudnessRemove background noise from audioRemove silence from audioConvert WAV to MP3
Cutting a ringtone, a clip or an intro
- Find the start and end by listening, and write them as minutes and seconds (1:12 to 1:42).
- Keep a little room at the end, because a cut on a frame can clip a final syllable by a few milliseconds.
- For a podcast intro, cut from the original episode file rather than a compressed copy.
- To add fades, cut here and fade in an editor; this tool does not fade.
Questions people ask
How do I cut an MP3 without losing quality?
Cut an MP3 with this tool: the frames are copied, so the sound is not encoded again. Our 129 kbps test file stayed at 129 kbps with identical samples.
How accurate is the cut?
Within one MP3 frame, 26.1 ms. Our 5 second start landed 15.5 ms away.
What happens to a WAV or M4A file?
It is saved as an MP3 at LAME quality 2. Our WAV excerpt came out at 101 kbps mono.
Can I keep only the first seconds of a file?
Yes. Set the start to 0 and the end to the length you want.
Can I add a fade in or out?
No. The tool cuts and does not process the sound.
What is the biggest file I can cut?
100 MB. The first run downloads an engine of about 30 MB once per visit.
Is my audio uploaded?
No. The file is read and cut in your browser.
Related
Sources and how we checked
- FFmpeg Documentation: ffmpeg (read 2026-10-09)
- FFmpeg wiki: Seeking (read 2026-10-09)
Reviewed October 9, 2026. The measurements ran the tool’s own ffmpeg arguments on the same single-thread WebAssembly engine your browser downloads. The script, the results and the saved documentation are in data/seo/hub/evidence/2026-10-09/tools.