A quote from an interview. You have fifty minutes of conversation and need the forty seconds in which the guest says the thing worth sharing. Find the passage, set the two points and the clip is ready for a radio piece, a social post or a presentation.
Dead air at the start. Recordings rarely begin with the first word: there is the fumbling with the recorder, the wait for latecomers, the "can everyone hear me". Set the start to the moment the talk begins, leave the end where the file ends, and the listener is spared the first three minutes.
A ringtone or notification sound. Pick the twenty or thirty seconds of the song you like and save them. Android phones accept the MP3 directly as a ringtone; an iPhone needs it converted to its own ringtone format in GarageBand or on a computer, which this tool does not do.
A clip for a voice message. Instead of forwarding a whole lecture or a ten-minute memo, cut the part the other person needs and send that. A minute of MP3 is around a megabyte and goes through any messenger.
The cutter works by time, not by dragging over a waveform. Play the recording, pause a moment before the passage starts and press "Set start to current position"; do the same for the end with "Set end to current position". Then play the selection. If a word is clipped at the front, move the start back by half a second and listen again.
A practical habit is to leave a short breath of silence on both sides. The tool does not add fades, so a cut made in the middle of a sound begins and ends abruptly, while a cut placed in a pause between sentences sounds natural without any further processing.
It does one thing: it saves a single range of a file as an MP3. Each run produces one clip. To get three excerpts, cut three times. To remove a section from the middle you would need to cut the part before it and the part after it, and joining those two pieces again requires an audio editor, because this tool does not merge files.
There are no fades, no volume changes, no noise removal and no waveform view, and the output is always MP3. Since the clip is encoded anew, trimming an MP3 or another lossy file loses a small amount of quality; at 192 kbps or above this is not something you will notice in practice.
The page runs ffmpeg as a WebAssembly program in your browser. The recording is opened from your disk, the range is encoded in memory and the clip is saved back to your device. No server receives the file, which matters when the source is a confidential interview or a private conversation.
On first use the engine, about 30 MB, is downloaded and cached. Audio files of any ordinary length are handled on phones and computers alike. If the source is a video of several gigabytes, a phone may run out of memory, and a computer is the better place to cut it. Protected files, such as songs downloaded inside a streaming app, are encrypted and cannot be opened.
A quote usually has to be written down as well as heard. After the cut, the "Transcribe this file" button sends the clip to Transcrify's free transcription. Unlike the trimming, that step uploads the MP3 to Transcrify's own servers, where the speech is recognised without involving a third-party service.
A short clip comes back as a complete transcript with timestamps, even without an account. 10 recordings are free, each up to 3 hours and 5 hours in total, the same allowance for guests and free accounts. Cutting first is also a way to spend that allowance sensibly: transcribe the twenty minutes you need rather than the two hours around them.