SILENTIQ is an AI noise-suppression plugin (VST3 / AU).
It removes noise from voice tracks of all kinds — home-recorded vocals, podcasts, narration and singing — in real time, using only the CPU.

What’s new in v1.7: A new screen. Pick the MODE on the left and compare the spectrum before (grey) and after (cyan) processing on the right. The screen now also comes in English (when the OS display language is not Japanese, the screen is shown in English). Hover over a control to see what it does in the bar at the top.

What’s new in v1.6: Fixed a small timing offset in DAW latency compensation (PDC). Timing now lines up at sample rates other than 48 kHz (such as 44.1 kHz) and when switching BYPASS.

1. Quick start

1. Insert SILENTIQ on the track you want to clean up.
2. Choose a MODE for the material (NARRATION for speech, VOCAL for singing).
3. Play the audio.
4. Check that BYPASS at the top right is off (while it is lit, you hear the original).
5. Set the amount of noise removal with STRENGTH.
6. If the voice sounds unnatural, raise VOICE PROTECT.

2. The screen

Help bar (top)

• Shows what the control under the mouse pointer changes.
• The screen is shown in Japanese when the OS display language is Japanese, and in English otherwise.

MODE (left)

• Two buttons switch between NARRATION (speech) and VOCAL (singing). The selected one lights up in cyan.
• The mode changes as soon as you click. While switching, the display changes briefly; once ready, processing continues with the new model (the sound does not drop out).
• Switching modes keeps your STRENGTH / VOICE PROTECT settings.

Spectrum display

• Overlays the spectrum before processing (grey area) and after processing (cyan line) in real time.
• When noise removal is working, the cyan line sits below the grey area where there is no voice.
• Nothing is drawn while there is no audio.

Knobs (below the spectrum)

• STRENGTH — amount of noise removal (0–100%).
• VOICE PROTECT — keeps the voice natural (0–100%).
• Double-click a knob to type in a value.

BYPASS (top right)

• While lit, processing stops and the original sound passes through. Use it to compare before and after.

3. Controls in detail

MODE (model)

Switches the noise-removal model to suit the material.

Mode	Use	Character
NARRATION (default)	Speech of all kinds narration / podcasts / streams / meetings	Removes noise from speech strongly.
VOCAL	Singing	A model suited to singing. It is gentle on held notes, vibrato and overtones, and keeps the song natural while reducing noise.

• If unsure, choose NARRATION for speech and VOCAL for singing.
• Applying NARRATION strongly to singing can shave off held notes and vibrato and make the voice sound unnatural. If that happens, switch to VOCAL.

STRENGTH (amount of noise removal)

Sets how strongly the selected model suppresses noise.

• Higher values remove noise more aggressively.
• Too high can make the voice sound unnatural or degrade the sound.
• A steady hiss (white noise) usually disappears at 20–40%.
• Default: 75%

VOICE PROTECT

Sets how strongly the voice is protected.

• Higher values protect more of the voice.
• If noise remains, raise STRENGTH and raise this as well.
• Default: 85%

BYPASS

• Off: noise removal is active.
• Lit: bypassed. The original sound passes through.
• A 20 ms crossfade is applied when switching to prevent clicks.

4. Suggested settings

Narration / podcasts (standard)

MODE:          NARRATION
STRENGTH:      75%
VOICE PROTECT: 85%

Noisy rooms (air conditioner, extractor fan, etc.)

MODE:          NARRATION
STRENGTH:      90%
VOICE PROTECT: 80%

White noise (a steady hiss)

MODE:          NARRATION
STRENGTH:      20–40%
VOICE PROTECT: 85%

Light hiss only

MODE:          NARRATION
STRENGTH:      50%
VOICE PROTECT: 90%

Singing (home-recorded vocals)

MODE:          VOCAL
STRENGTH:      75%
VOICE PROTECT: 85%

5. Latency

SILENTIQ adds a fixed latency for processing.
DAWs with latency compensation (PDC) correct it automatically.

MODE	Latency (samples)	Latency (time @48kHz)
NARRATION	1920 samples	about 40 ms
VOCAL	2880 samples	about 60 ms

The latency depends on the MODE. When you switch, the DAW’s latency compensation (PDC) follows automatically. At sample rates other than 48 kHz the time is almost the same and is compensated automatically.

6. Requirements

Item	Details
OS	Windows 10 / 11 x64 macOS 11 or later (Apple Silicon / Intel)
Format	Windows: VST3 / Mac: VST3, AU
DAW	DAWs that support VST3 / AU (Cubase, Studio One, Ableton Live, Reaper, Logic Pro, etc.)
Sample rate	44.1 kHz / 48 kHz (other rates are resampled internally)
CPU	Windows: x86_64 / Mac: Apple Silicon, Intel (no GPU needed)
Channels	Stereo / Mono
Screen language	Japanese / English (follows the OS display language)

7. Installing and uninstalling

Installing (Windows)

1. Extract the ZIP.
2. Copy the SILENTIQ.vst3 folder to:

   C:\Program Files\Common Files\VST3\

3. Rescan VST3 plugins in your DAW.

Installing (Mac)

1. Extract the ZIP.
2. Copy SILENTIQ.vst3 and SILENTIQ.component to their locations (~ is your home folder. In Finder, open “Go” → “Go to Folder”, paste the path, and open it. Create the folder if it does not exist).

   SILENTIQ.vst3      → ~/Library/Audio/Plug-Ins/VST3/
   SILENTIQ.component → ~/Library/Audio/Plug-Ins/Components/

3. Open Terminal (Applications → Utilities → Terminal), paste the following three lines one by one, and press Return.

   cd ~/Library/Audio/Plug-Ins
   xattr -dr com.apple.quarantine "VST3/SILENTIQ.vst3"
   xattr -dr com.apple.quarantine "Components/SILENTIQ.component"

   If nothing is displayed, it succeeded. For the trial, read SILENTIQ as SILENTIQ Trial.

4. Rescan plugins in your DAW (if the AU does not appear in Logic Pro, restart your Mac).

Uninstalling

On Windows, delete the following folder.

C:\Program Files\Common Files\VST3\SILENTIQ.vst3

On Mac, delete the SILENTIQ.vst3 and SILENTIQ.component you copied.

Upgrading from the trial

1. Delete the trial.

   Windows: C:\Program Files\Common Files\VST3\SILENTIQ Trial.vst3
   Mac:     ~/Library/Audio/Plug-Ins/VST3/SILENTIQ Trial.vst3
            ~/Library/Audio/Plug-Ins/Components/SILENTIQ Trial.component

2. Install the full version (steps above).
3. Rescan plugins in your DAW.
4. In your DAW projects, replace SILENTIQ Trial with SILENTIQ.

8. Troubleshooting

• Noise is not removed
  • Check that BYPASS is off.
  • Check that STRENGTH is not too low.
• The voice sounds unnatural or degraded
  • Lower STRENGTH.
  • Raise VOICE PROTECT.
• The plugin does not appear in the DAW
  • Check that you copied it to the right plugin folder.
  • Rescan plugins in your DAW.
  • Windows: check that df.dll and onnxruntime_siq.dll are inside the SILENTIQ.vst3 folder.
  • Mac: if you see “is damaged and cannot be opened”, or the plugin is not listed, run the commands in step 3 of the installation.
• VOCAL mode sounds no different, or the same as NARRATION
  • Using the plugin without extracting the ZIP can leave DLLs missing. Copy the whole folder to the VST3 folder again.
  • Check that onnxruntime_siq.dll (the engine for the VOCAL model) is inside the SILENTIQ.vst3 folder.
• No sound
  • Sound comes out even while BYPASS is lit (the original). If there is no sound, check the track’s output and mute in your DAW.
  • Check your DAW’s buffer size (any size is supported).
• Security software reacts
  • df.dll (the AI processing engine) may be flagged by mistake.
  • Add an exception, or release it from your security software’s quarantine.

9. Notes

• This product is provided “AS IS”, without support.
• No refunds will be provided.
• Source code is not public.
• Copyright of this product belongs to Tonecraft Tokyo.
• Redistribution, resale, modification, and reverse engineering of this product are prohibited.
• Licensed for use by the purchaser only. Transfer or sharing with third parties is prohibited.
• This product uses DeepFilterNet3 / DeepFilterNet2 (Apache License 2.0), ONNX Runtime (MIT License), and Noto Sans JP / JetBrains Mono (SIL Open Font License 1.1).

10. Disclaimer

• This product is provided “AS IS”, without warranty of any kind, express or implied, including but not limited to merchantability, fitness for a particular purpose, and non-infringement of third-party rights.
• Tonecraft Tokyo is not liable for any damages arising from the use of, or inability to use, this product, including but not limited to loss or corruption of project files or audio data, equipment malfunction, lost working time, or lost profits.
• Operation is not guaranteed for every combination of DAW, operating system, and hardware.
• Specifications may change, and the product may be discontinued, without notice.
• Use this product at your own risk. Back up your projects before important work.