← All plugins

Voice isolation · VST3 + Standalone

VoxIsolate

Pull a clean voice out of anything.

Hybrid voice isolation that combines a neural network with classic signal processing. RNNoise removes keyboards, traffic and chatter, an OM-LSA stage removes steady hiss and room tone, and an adaptive De-Hum subtracts mains hum and fan tones.

Instant download · secure checkout by Stripe, billed as EssentialApps · the bundle saves $22

Version 1.1.1 · Windows 10/11, 64-bit · needs the VC++ runtime · already bought it? Find your downloads

VoxIsolate AIISOLATION REDUCTION DE-HUM LOW CUT PRESENCE MIX OUTPUT

Each engine handles the noise the other can't. The recurrent network catches non-stationary noise that classic DSP misses. The classic stage tracks the noise floor continuously, so you never have to capture a noise profile, and removes whatever steady noise is left.

De-Hum finds steady tones (50/60 Hz mains hum and buzz, fan and HVAC tones at any pitch) and subtracts them sample by sample instead of notching, so the voice's own harmonics pass through untouched. It does nothing until a hum is confirmed. Mains hum is recognized in about a second; any other tone has to hold steady for eight seconds and still be there in the pauses between words, so a held note in background music or a sustained voice pitch isn't mistaken for hum.

Every setting was calibrated on real speech in seven noise types and checked on voices that were never used for tuning. Against version 1.0 the defaults are cleaner (SI-SDR +6.5 dB vs +3.8 dB) and easier to understand (intelligibility now improves instead of dipping), with the same noise removal and less musical noise. Under mains hum, De-Hum alone takes about 27 dB of hum off a voice.

Features

  • AI Isolation

    Xiph's RNNoise recurrent network, running internally at 48 kHz.

  • Classical reduction

    OM-LSA gain with continuous noise tracking. No noise-profile step needed.

  • De-Hum

    Adaptive hum and tone subtraction for mains and HVAC, up to two hums at once, with a readout of what it has locked onto.

  • Stereo-linked

    One suppression decision for both channels, so a stereo voice stays where it was in the image.

  • Phase-aligned Mix

    Parallel processing without comb filtering, because the dry path is latency-compensated.

  • Exact latency

    Measured at load with an impulse and reported to the host, so it stays sample-accurate.

Factory presets

Factory presets

  • Podcast Cleanup
  • Streaming / Live Voice
  • Video Call / Laptop Mic
  • Voiceover Studio Polish
  • Field Recording Rescue
  • Music Bleed Removal
  • Gentle Denoise (Preserve Air)
  • Aggressive Rescue
  • Parallel Blend (50%)

Presets also appear in your DAW's own factory-preset browser.

Install

  1. Buy and download the zip from your library. The link is emailed to you too.
  2. Unzip it anywhere.
  3. VST3: copy the VoxIsolate.vst3 folder to C:\Program Files\Common Files\VST3, then rescan plugins in your DAW.
  4. Standalone: run VoxIsolate.exe and choose your input and output under Options → Audio Settings. Use headphones when monitoring a live mic.

Windows SmartScreen may warn about a new publisher on first launch. Choose More info → Run anyway. Why this happens.