VoxCut is a dedicated silence remover for audio and video: it automatically detects and cuts silent pauses — dead air, long breaths, gaps between takes — from podcasts, interviews and recordings. Unlike full editors that bundle silence removal as one feature among many, it does this one job, entirely on your device.
I want to walk through the actual engineering problem behind it, because "remove silence from audio" sounds trivial until someone hands you a 72-hour, 5GB recording and expects it back in a few minutes.
The naive approach falls over first
The obvious way to build this: load the whole file into memory, scan the waveform, cut the quiet parts, re-encode. That's roughly what most desktop audio editors do — and it's exactly why they choke on anything past a couple of GB. Once you're loading multi-hour recordings, you hit a RAM ceiling that has nothing to do with your CPU.
VoxCut instead runs a local, FFmpeg-based streaming pass over the file rather than materializing the whole thing in memory. That single architectural choice is what lets it handle files "beyond 5GB" without a hard ceiling — the real limit becomes your disk space, not your RAM.
Detection: comparing amplitude against a threshold
The detection logic itself is simple in concept and easy to get wrong in practice: compare each segment's amplitude against a decibel threshold and a minimum pause duration. Too aggressive, and you cut a natural breath mid-sentence. Too conservative, and you leave dead air in.
VoxCut previews exactly what will be cut before anything is removed, so the sensitivity setting is something you tune by ear rather than trust blindly — nothing is deleted until you confirm.
The Turbo Mode trade-off (V2)
The 2.0 release added a Pro-only Turbo Mode that's a good example of an honest trade-off rather than a free performance win. The standard export stitches kept segments together with crossfades, which sounds smoother but costs time on files with hundreds of detected cuts. Turbo mode removes silence in a single pass instead — no crossfades — which is roughly 2x faster on realistic content (podcasts, interviews, meetings).
The cost: sharper cuts instead of smooth transitions. For most spoken-word content that's barely audible; for content that leans on the crossfades to hide the cut points, standard mode is still the default.
On realistic content (not the "silence-only" edge case that inflates marketing numbers), expect roughly 35-50x real-time in standard mode and 70-100x in Turbo, depending on the machine. A 72-hour, 5GB file was processed in under 6 minutes on ordinary consumer hardware.
Going cross-platform: Windows/Linux desktop, then Android
VoxCut started as a Windows/Linux desktop app. Porting the same core logic to Android (shipped as v2.0.2, same week as the 2.0 desktop release) meant a separate app with its own constraints: mobile storage, a different free-tier limit (5 minutes per file vs. 10 minutes on desktop), and Pro features gated differently — AI noise reduction and loudness normalization are Android-Pro-only, priced independently (€14.99 lifetime) from the desktop license.
Keeping "100% local, no upload" true on both platforms was the one non-negotiable constraint across the port — it's the core promise, not a checkbox feature.
What I'd do differently
If I started over, I'd build the streaming-pass architecture from day one instead of retrofitting it — the in-memory prototype worked fine on my own test files and only broke once real users started throwing multi-hour recordings at it. Test with files bigger than you think anyone will use.
VoxCut is free to try (Windows, Linux, Android) — voxcutpro.com. Feedback and questions welcome in the comments, this is my first shipped piece of software and I'm still learning in public.
About VoxCut
VoxCut is a dedicated silence remover for audio and video. It automatically removes silence, pauses and dead air while processing your files locally — with no upload and no subscription.
VoxCut — Silence Remover for Audio & Video · https://voxcutpro.com
Top comments (0)