AI VOICE DENOISER · VST3 · WINDOWS

Voice on one knob.
The room on the other.

minoise uses on-device AI to split a recording into speech and background, then lets you choose how much of each to keep. Free, offline, and ready to drop into your DAW.

Free download for Windows v2.0.1 · VST3 · 64-bit · ~12 MB · Quark Drive
Hear it first
minoise plugin window with Voice at 140% and Ambience at 35%, showing the three-layer spectrum and output readout
Signal-to-noise
5.0 rising to 10.4 dB
Latency, auto-compensated
≈ 38 ms
One core (Standard)
≈ 10%
Runs locally, no internet
Offline

LISTEN

Hear it. Then turn it.

This is the plugin's actual output. The page mixes it live with the same formula the plugin uses:
Output = Voice × denoised speech + Ambience × removed background.

Processed locally
SPECTRUM
Voice Ambience Input
Your browser can't draw the spectrum.

Press play to start

OUTPUT / INPUT Output vs. input — dB Live RMS ratio
Sample A Audio loads when you press play
Compare

Voice

Denoised speech

GAIN ±0.0 dB Double-click for 100%

Ambience

Removed background

GAIN ±0.0 dB Double-click for 100%
Presets

Drag up/down, scroll or use arrow keys · PageUp / PageDown ±5 · Home / End to either end

Sample A is 48 kHz stereo. Samples B and C are 16 kHz mono.

TWO KNOBS

Noise reduction, as a mix.

Pick a balance. One click and you'll hear it in the listening desk.

MEASURED

The noise goes. The highs stay.

Signal-to-noise

0–12 dB
SNR gain+5.4 dB
Above 8 kHz−2 / +1.8 dB

Deviation from the clean original. No muffled top end.

+20−28k12k16k20k Illustrative

Test material: a noisy 48 kHz voice recording.

MEASURED · PERFORMANCE

Light enough for every track.

Standard

STANDARD
8–10×real time

about 10% of one core

Lite

LITE
15–19×real time

about 5% of one core

01

In Lite mode, stereo runs the model once, roughly halving CPU.
Switch modes during playback without pops.

Latency

48 kHz
≈ 37.9 ms
44.1 kHz
≈ 37.9 ms
16 kHz
33.25 ms

Latency is reported to your DAW for automatic delay compensation.

Measured at 48 kHz stereo on a single core.

INTERFACE

Every move, on screen.

A touch of room: minoise with Voice at 140% and Ambience at 35%
  1. 1

    Live spectrum

    Input, voice and ambience, layered in one view.

  2. 2

    Reduction

    A readout plus a trend line of the last 4 seconds.

  3. 3

    LED-ring knobs

    5% per step. Double-click to return to default.

  4. 4

    Level meters

    Input and output, with peak hold.

  5. 5

    Standard / Lite

    Switch any time during playback.

  6. 6

    Bypass

    One click to hear the original.

INSTALL

Installed in three steps.

Installer: minoise-2.0.1-win64-vst3.zip, about 12 MB.

  1. 01

    Download and unzip

    Unzip minoise-2.0.1-win64-vst3.zip.

  2. 02

    Into the VST3 folder

    Move the minoise.vst3 folder into:

    C:\Program Files\Common Files\VST3
  3. 03

    Rescan plugins

    Rescan plugins in your DAW.

System requirementsWindows 10 / 11 · 64-bit · VST3-compatible host

Free download for Windowsv2.0.1 · VST3 · 64-bit · ~12 MB · Quark Drive

FAQ

Questions.

Is it free?

Yes.

Does it need an internet connection?

No. The model is built into the plugin, and your audio stays on your machine.

Will it make voices sound muffled?

Above 8 kHz, it stays within −2 / +1.8 dB of the clean original.

How much latency does it add?

About 37.9 ms at 48 kHz. The plugin reports it to your DAW, which lines everything up automatically.

Standard or Lite?

Lite runs the model once for stereo, roughly halving CPU. Reach for it in big sessions or when the machine is busy. You can switch any time during playback.

Can I export the ambience on its own?

Turn Voice down to 0% and leave Ambience at 100%. The output is ambience only.

MINOISE

Your audio stays on your machine.

minoise · Free · Windows VST3

Free download