Word-accurate voice slicing for Mac & Windows
Cut audio by selecting text.
Highlight a phrase and the waveform jumps straight to it. Export it as a named, ready-to-deliver clip. Your audio never leaves your device.
Every feature, free for seven days. No account, no card. Also for macOS. Also for Windows.
How it works
Three steps, from take to delivery.
Load a recording, select the words you want, export the clips. It all runs locally, so transcribing still works with the wifi off.
1. Load your audio
Open a WAV, MP3, FLAC, M4A, AAC or OGG file, of any length. Vocal Slice transcribes it locally with Whisper, producing timestamps down to the individual word.
2. Select the words
Highlight a phrase in the transcript. The waveform zooms to exactly that region, with draggable start and end handles for frame-accurate trimming.
3. Export named files
Preview on loop, then export. A WAV source is cut byte-perfect, so what you hand over is the original audio rather than a re-encode of it, and every file is named from your own template.
Features
Built for people who cut voice recordings all day.
Podcasters, video editors, voiceover artists, content creators: anyone who works through hours of recordings to find the few parts that matter.
Named to your convention
Every slice is written from your own template: {source}_{index}_{slug}, timestamps, or whatever your project expects. A session's worth of clips comes out already matching the convention you deliver in, rather than needing an evening of renaming first.
Source-quality exports
WAV files are sliced losslessly at the byte level, so a slice carries exactly the source's channel count, sample rate and bit depth. Stereo and multichannel come through untouched. Other formats decode to 24-bit WAV.
Re-trim without starting over
Open any slice back up in the list and move its boundaries, either by dragging its own waveform or by typing exact start and end times. The transcription is still loaded, so a second pass costs you nothing but the adjustment itself.
Speaks your language
English and multilingual Whisper models, from fast and light to slow and accurate, with languages from Spanish and German through to Japanese, Arabic and Hindi.
Find every take of a line
Select a phrase and Vocal Slice finds every other place it occurs in the recording, then steps between them with the waveform following each one — five takes of the same line, ready to compare. The search box works the same way for any word, with a running match count.
Nothing leaves your machine
Transcription and slicing both happen on your own hardware. There's no account to create, and your audio is never sent anywhere. That's what makes it usable for unreleased episodes, dialogue under NDA, or an interview you promised to protect.
Pricing
Buy it once. It's yours offline.
The seven-day trial comes first, and doesn't ask for a card. After that: every feature and every future version, on up to three machines.
- Every feature. Transcription, text-selection slicing and lossless export.
- Updates, forever. Every new version is included.
- Three machines. Desktop, laptop, and one more.
- Seven-day trial. Everything unlocked, before you pay a penny.
Runs entirely on your machine. Your audio never leaves it. Vocal Slice checks your licence once, when you activate it, and then doesn't need us at all — it keeps working offline, indefinitely, on a machine that never sees the internet again.
Secure checkout via Polar. Tax included.
Requirements
What you'll need.
- Operating system
- Windows 10 or 11 (64-bit), or macOS 11 Big Sur and later
- Graphics
- Any WebGPU-capable GPU, including the one in every Apple Silicon Mac and most modern integrated graphics. CPU fallback included
- Disk space
- ~200 MB for the app, plus ~120 MB–1.6 GB per transcription model
- Internet
- Needed once to download a model and activate your licence. Not for transcribing
Questions
The things people ask first.
Isn't the transcription just Whisper, which is free?
The models are OpenAI's, MIT-licensed, and anyone can run them. What you're paying for is everything between a transcript and a delivered file: correcting word timings that arrive late, coupling the transcript and waveform so each drives the other, filename templating, and byte-level slicing that cuts from the source rather than re-encoding it.
How accurate are the cut points?
Word timings from speech recognition run consistently late, so Vocal Slice shifts them earlier by a compensation that defaults to 300ms and can be changed in settings. Both boundaries move together, which preserves each word's duration, and the correction applies to the exported cut rather than only to what's displayed.
Does it work without an internet connection?
Transcribing never needs one. You need a connection twice: once to download a transcription model, and once to activate your licence. After that it runs with the network off, which is the point on a plane or in a studio with no guest wifi.
What audio can it open?
WAV, MP3, FLAC, M4A, AAC and OGG, of any length. Clips come out as WAV: a WAV source is cut byte-exact at its original sample rate, bit depth and channel count, and the other formats decode once to 24-bit rather than being re-compressed per clip.
Is my audio ever uploaded?
No. Transcription and slicing both run on your own hardware, and there's no account to create. That's what makes it usable for unreleased episodes, dialogue under NDA, or an interview you promised to protect.
Is it really a one-time payment?
Yes. You pay $29 once and Vocal Slice is yours, including every future version. Once your key is activated the app doesn't need to check in again, so it keeps working even if it never sees the internet after that.
Download
Try it on your own audio.
Free for seven days with every feature available. Runs on Windows and macOS, Apple Silicon and Intel alike.
Windows 10/11 64-bit, macOS 11 and later (Apple Silicon and Intel). Full requirements.
Windows may warn you the first time. Vocal Slice isn't yet signed with a code-signing certificate, so SmartScreen shows “Windows protected your PC — unknown publisher”. Choose More info → Run anyway to continue.