Documentation

Kiyome User Manual

Complete guide to AI-assisted de-mudding — cleaner mixes, one knob at a time.

Version 1.0.0

EN 繁體中文 简体中文 日本語 한국어

1. Introduction

What is Kiyome?

Kiyome (清め) is an AI-assisted de-mudding plugin that detects and removes muddy frequency buildup in the critical 100–700 Hz range. Using spectral pattern recognition powered by an embedded ONNX neural network, Kiyome intelligently identifies problematic low-mid energy and surgically reduces it — preserving warmth, body, and musicality.

AI-powered clarity meets one-knob simplicity. Select your instrument, turn the Depth knob, and let the neural network handle the rest. Or dive deeper with 5-band parametric control, M/S processing, and three analog color circuits for character and warmth.

Key Features

  • ✓
    One-Knob Operation (Depth)

    A single master control scales all processing — from subtle cleanup to aggressive de-mudding. Perfect for fast workflows.

  • ✓
    13 Instrument-Specific AI Models

    Dedicated neural network models for Male Vocal, Female Vocal, Kick, Snare, Toms, Overheads, Electric Bass, Synth Bass, Acoustic Guitar, Electric Guitar, Keys, Strings/Brass, and Full Mix.

  • ✓
    Learn Mode

    Analyzes your audio's spectral profile in 5 seconds and automatically adjusts per-band sensitivity for optimal de-mudding.

  • ✓
    5-Band Parametric EQ (100–700 Hz)

    Five precisely tuned bands targeting Sub-Mud (100 Hz), Low-Mud (200 Hz), Core-Mud (350 Hz), Upper-Mud (500 Hz), and Mud-Warmth (700 Hz).

  • ✓
    M/S Processing

    Process mid and side channels independently for surgical stereo-field de-mudding without collapsing the mix.

  • ✓
    Analog Color (Neve / API / Studer)

    Three independent analog character circuits — Warm (Neve 1073), Punch (API 2520), and Tape (Studer A800) — to add life back after de-mudding.

  • ✓
    38 Factory Presets

    Professionally crafted presets across 10 categories covering vocals, drums, bass, guitar, keys, strings, mix bus, genre-specific, creative, and analog color starting points.

  • ✓
    Real-Time Spectral Visualization

    See exactly where mud lives in your audio with a live spectrum display (60–1000 Hz), draggable band nodes, and a gain reduction curve overlay.

Who Is It For?

Whether you are a mixing engineer dealing with boxy vocals and cloudy guitars, a producer trying to get a clean low-mid balance, a podcaster fighting room resonance, or a content creator who wants clearer audio without spending hours with an EQ — Kiyome gives you AI-powered clarity in seconds.

2. System Requirements

🍎 macOS

  • OS Version: macOS 10.15 Catalina or later
  • Processor: Apple Silicon (M1/M2/M3/M4) & Intel (Universal Binary)
  • Plugin Formats: AU (Audio Unit) + VST3
  • Standalone: Standalone application included
  • DAW Compatibility: Any DAW supporting AU or VST3 (Logic Pro, Ableton Live, Pro Tools, FL Studio, Reaper, Cubase, Studio One, GarageBand, etc.)

🖥 Windows

  • OS Version: Windows 10 or later
  • Processor: 64-bit (x86_64)
  • Plugin Format: VST3
  • DAW Compatibility: Any DAW supporting VST3 (Ableton Live, FL Studio, Reaper, Cubase, Studio One, Pro Tools, etc.)

Note: The embedded ONNX neural network runs on CPU — no GPU required.

3. Installation Guide

🍎 macOS

  1. 1. Download the Kiyome installer (.pkg) from sugoiaudio.com.
  2. 2. Double-click the .pkg file and follow the on-screen instructions.
  3. 3. The plugin will automatically install to the correct locations.
  4. 4. Open your DAW and rescan plugins if needed. The plugin appears as “Kiyome” by “Sugoi.”
  5. 5. Your 30-day free trial starts automatically — full functionality, no limitations.

Installation Paths:

AU: ~/Library/Audio/Plug-Ins/Components/Kiyome.component VST3: ~/Library/Audio/Plug-Ins/VST3/Kiyome.vst3

Tip: If the plugin does not appear in your DAW after installation, try rescanning your plugin folder. In Logic Pro, this happens automatically on launch. In Ableton Live, go to Preferences > Plug-ins > Rescan.

4. Interface Overview

Kiyome features a clean, focused interface designed for fast de-mudding workflows. The layout is organized into distinct zones: toolbar, spectrum display, knob panel, secondary controls, status bar, and an expandable advanced panel.

Toolbar

The top bar contains the following controls from left to right:

  • Brand Logo: Sugoi Audio branding.
  • Preset Navigation: Previous/next arrows with a dropdown menu for browsing and loading presets.
  • Undo / Redo: Step backward or forward through your parameter change history.
  • A/B Compare + Copy: Toggle between two parameter snapshots (A and B). Copy syncs both slots to the current settings.
  • Bypass: Bypass the entire plugin processing.
  • Info (How to Use): Opens a quick-start guide overlay.
  • Settings: Access licensing, activation, and advanced options.

Spectrum Display

The large display area shows a real-time spectral visualization of the 60–1000 Hz range:

  • 5 Draggable Band Nodes: Each node represents one of the five de-mudding bands. Click a node to select that band; drag vertically to adjust its sensitivity.
  • Gain Reduction Curve: A real-time curve (in dB) showing how much reduction is being applied at each frequency.
  • Input Magnitude Overlay: Shows the incoming signal's spectral energy distribution so you can see where mud accumulates.
  • Right-Click: Right-click any band node to bypass that individual band.

Knob Panel

Three rotary knobs arranged across the center of the interface:

Depth

110px, terracotta. Master processing strength (0–100%).

Speed

68px, teal. Reaction time (0–100%).

Tone

68px, grey-brown. EQ bandwidth (0–100%).

All knobs respond to vertical drag: up to increase, down to decrease. Hold Shift for fine adjustment. Double-click to reset to default.

Secondary Controls

  • Instrument Selector: A dropdown menu with 13 AI models. Select the instrument type that matches your source material for optimal de-mudding behavior.
  • Learn Button: Initiates a 5-second spectral analysis of your audio. After learning, per-band sensitivities are automatically adjusted. See Section 6: Learn Mode for details.

Status Bar

A horizontal control strip below the knob panel:

  • Low Cut Slider: High-pass filter from 20–200 Hz with a slope selector (6, 12, 18, or 24 dB/oct).
  • Delta Listen: Toggle to hear only the frequencies being removed — invaluable for verifying you are cutting mud, not body.
  • Auto Gain: Toggle automatic output level compensation to match the perceived loudness of the bypassed signal.
  • Advanced Panel Toggle: Opens/closes the Advanced panel. An orange dot indicator appears when any Analog Color parameter is active, even when the panel is closed.

Advanced Panel (Expandable)

Revealed by clicking the Advanced toggle. Contains three independent analog character sliders:

Warm

Neve 1073 character (0–100%)

Punch

API 2520 character (0–100%)

Tape

Studer A800 character (0–100%)

See Section 9: Analog Color for full descriptions of each circuit.

Footer Area

  • Processing Mode Selector: Switch between L/R (standard stereo) and M/S (Mid/Side) processing.
  • Stereo Link Slider (L/R mode): Controls how linked the left and right channels are (0% = dual mono, 100% = fully linked).
  • M/S Balance Slider (M/S mode): Controls the balance of de-mudding depth between the mid and side channels.

5. Core Parameters Reference

Parameter Range Default Description
Depth 0 – 100% 40% Master processing strength. Scales all band gain reductions proportionally. 0% = no processing, 100% = maximum de-mudding.
Speed 0 – 100% 50% Reaction time of the spectral detection. 0% = 200 ms smooth (slow, gentle), 100% = 10 ms (fast, reactive). Lower values are more transparent; higher values catch transient mud.
Tone 0 – 100% 50% Controls the EQ bandwidth of each band. 0% = narrow surgical cuts (Q × 1.5), 100% = wide aggressive cuts (Q × 0.5).
Output -12 to +6 dB 0 dB Post-processing level trim. Use to match the output level to the bypassed signal.
Instrument 13 models (default: Full Mix) Selects the AI model optimized for the source material. Each model has different spectral sensitivity profiles.

Available Instrument Models

Male Vocal
Female Vocal
Kick
Snare
Toms
Overheads
Electric Bass
Synth Bass
Acoustic Guitar
Electric Guitar
Keys
Strings / Brass
Full Mix

Parameter Interaction Notes

Depth + Instrument

The AI model determines which frequencies are targeted; Depth controls how aggressively they are reduced. Changing the instrument model can dramatically change the result even at the same Depth setting.

Speed + Tone

Speed controls reaction time (temporal), Tone controls bandwidth (spectral). For transparent processing, use lower Speed and moderate Tone. For aggressive cleanup, use higher Speed and higher Tone.

Auto Gain

Compensates for level changes caused by de-mudding. Disable Auto Gain when level-matching by ear, or when the plugin is followed by a compressor that reacts to level changes.

Delta Listen

When enabled, you hear only the frequencies being removed. If you hear musical content (not just muddiness), reduce the Depth or adjust band sensitivities to preserve it.

6. Learn Mode

What It Does

Learn mode analyzes the spectral energy distribution of your audio and automatically adjusts each band's sensitivity to match the material. Bands where mud is concentrated get higher sensitivity; bands where the signal is clean get lower sensitivity. This ensures Kiyome focuses its processing where it matters most.

How to Use

  1. 1. Play your audio through the plugin (ensure signal is audible, not silence).
  2. 2. Press the LEARN button. A 5-second countdown begins.
  3. 3. Keep the audio playing during the countdown (5s → 4s → 3s → 2s → 1s).
  4. 4. Learning stops automatically after 5 seconds. Band sensitivities are updated immediately.

Early Stop

You can stop the learning process early by clicking the button again during the countdown. Whatever has been analyzed so far will be used.

Clear Profile

Double-click the LEARNED button (green state) to clear the learned profile and reset all band sensitivities to their defaults.

Button States

State Appearance Meaning
LEARN Default / idle No profile active. Click to start learning.
5s / 4s / 3s / 2s / 1s Countdown display Actively analyzing audio. Click to stop early.
LEARNED Green indicator Profile is active. Band sensitivities have been adjusted. Double-click to clear.

Tip: The learned profile is saved with your presets. When you save a preset after learning, the profile is stored so you can recall it later without re-learning.

7. Band Controls

Kiyome uses five precisely tuned bands to target the most common muddy frequency regions. Each band can be individually adjusted for sensitivity and bypassed independently.

Band Frequency Character
Sub-Mud 100 Hz Deep rumble and sub-bass mud. Common in kick drums, bass, and room resonance.
Low-Mud 200 Hz Classic “boominess.” Proximity effect in vocals, acoustic guitar body resonance.
Core-Mud 350 Hz The “boxy” zone. Cardboard-like resonance in vocals, snare, and room recordings.
Upper-Mud 500 Hz Honky, nasal quality. Common in electric guitar, toms, and cheaper microphones.
Mud-Warmth 700 Hz Borderline between mud and warmth. Handle with care — too much reduction here can thin the sound.

Per-Band Sensitivity

Each band has an adjustable sensitivity ranging from 1.0x to 4.0x. Higher sensitivity means the band reacts more aggressively to detected mud. Adjust sensitivity by dragging the band nodes vertically on the spectrum display.

Per-Band Bypass

Right-click any band node on the spectrum display to toggle its bypass state. A bypassed band passes audio through unprocessed. The band detail panel (shown when a band is selected) displays the frequency, current gain reduction, sensitivity value, and bypass state.

Tip: If you find the plugin is removing too much body from a vocal, try bypassing the Mud-Warmth (700 Hz) band or reducing its sensitivity to 1.0x. This preserves the natural warmth while still cleaning the lower mud bands.

8. Processing Modes

L/R (Stereo)

Standard left/right stereo processing. Both channels are processed with the same de-mudding settings.

  • Stereo Link: 0% — Dual mono. Each channel is analyzed and processed independently. Maximum stereo separation.
  • Stereo Link: 100% — Fully linked. Both channels share the same detection signal, ensuring identical processing. Best for preserving stereo image.
  • Stereo Link: 50% (default) — A balanced compromise. Good for most material.

M/S (Mid/Side)

Mid/Side processing mode. The signal is decoded into mid (center) and side (stereo difference) components, each processed independently.

  • M/S Balance Slider: Controls the balance of de-mudding depth between mid and side. Center = equal processing. Move left to focus on mid; move right to focus on side.
  • Dual-Color Spectrum: In M/S mode, the spectrum display shows two overlaid traces — terracotta for the mid channel and purple for the side channel.

Pro tip: M/S mode is excellent for mix bus de-mudding. Process the mid channel more aggressively (where mud accumulates from centered instruments like bass and kick) while leaving the side channel cleaner to preserve stereo width and ambiance.

Low Cut (High-Pass Filter)

Frequency Range: 20 – 200 Hz

Slope Options: 6 dB/oct, 12 dB/oct, 18 dB/oct, 24 dB/oct

A high-pass filter applied before the de-mudding processing. Use it to remove sub-bass rumble and infrasonic content that can interfere with the AI's spectral analysis. For most material, 30–60 Hz at 12 dB/oct is a safe starting point.

9. Analog Color

De-mudding can sometimes leave audio feeling thin or clinical. The Analog Color section adds warmth, character, and harmonic richness back into the signal. All three circuits are independent and can be combined freely. An orange dot appears on the Advanced toggle when any color is active, even when the panel is closed.

Warm — Neve 1073

Range: 0 – 100%  |  Default: 0%

Odd-harmonic dominant saturation with subtle even harmonics and a warm low-pass filter centered around 300 Hz. Inspired by the legendary Neve 1073 preamp, this circuit adds a gentle, musical warmth that fills out the low-mids naturally.

Best For

  • • Vocals that sound thin after de-mudding
  • • Acoustic instruments that need body
  • • Adding richness without obvious distortion

Punch — API 2520

Range: 0 – 100%  |  Default: 0%

Cleaner and tighter than Warm. Features dual-envelope transient shaping with mild harmonic drive. Inspired by the API 2520 op-amp, this circuit adds punch and definition without the thickness of tube-style coloration.

Best For

  • • Drums and percussive material
  • • Electric bass that needs definition
  • • Adding transient snap and energy

Tape — Studer A800

Range: 0 – 100%  |  Default: 0%

Deep saturation with a natural high-frequency rolloff at 16 kHz. Inspired by the Studer A800 tape machine, this circuit adds vintage character, gentle compression, and a cohesive “glue” that makes everything sound like it belongs together.

Best For

  • • Mix bus processing for vintage character
  • • Taming harsh digital recordings
  • • Full mixes that need cohesion and depth

Tip: The three color circuits can be combined. A popular starting point is Warm at 30% + Tape at 20% for vocals — this replaces the body lost from de-mudding with a musical, analog-flavored warmth.

10. Factory Presets

Kiyome ships with 38 factory presets organized into 10 categories. Each preset configures the instrument model, Depth, Speed, Tone, band sensitivities, and Analog Color settings for common use cases.

Category Presets Examples
Vocals 5 Male vocal clarity, female vocal cleanup, podcast de-mud, vocal bus, narration
Drums 4 Kick tightener, snare clarity, toms cleanup, overhead de-mud
Bass 3 Electric bass focus, synth bass cleanup, upright bass clarity
Guitar 2 Acoustic guitar de-box, electric guitar tighten
Keys & Synth 4 Piano clarity, Rhodes warmth, synth pad cleanup, organ de-mud
Strings & Orchestral 3 String ensemble, brass section, orchestral bus
Mix Bus 4 Gentle mix cleanup, mastering touch, aggressive de-mud, M/S clarity
Genre 6 Pop mix, rock mix, hip-hop, EDM, jazz, acoustic/folk
Creative 3 Extreme scoop, lo-fi thinning, telephone effect
Color 4 Neve warmth, API punch, Studer tape, all-color blend

Preset Management

  • Save As: Save your current settings as a new user preset.
  • Delete: Remove a user preset (factory presets cannot be deleted).
  • Export: Export a preset to a file for sharing or backup.
  • Import: Import a preset file from disk.

A/B Compare

Toggle between two independent parameter snapshots (A and B) using the A/B button in the toolbar. Use Copy to sync both slots to the current settings — then tweak one slot to quickly compare variations.

Undo / Redo

Full parameter change history. Every knob turn, button click, and preset load is recorded. Use the Undo and Redo buttons in the toolbar to step through your changes.

11. Trial & Licensing

30-Day Free Trial

  • Full functionality during the trial period — no feature limitations.
  • All 13 AI models, all 38 presets, all parameters, all Analog Color circuits.
  • Trial starts automatically on first launch.
  • When the trial expires, audio output is muted. Purchase a license to continue using the plugin.

Purchasing a License

Visit the official website to purchase a license key. Payment is processed securely through Lemon Squeezy.

Buy Now

Online Activation

  1. 1. Open Kiyome and click the Settings icon in the toolbar.
  2. 2. Paste your license key in the activation field.
  3. 3. Click “Activate” to complete the activation.

Offline Activation

If your studio computer does not have internet access:

  1. 1. Open Settings and note your Machine ID.
  2. 2. On a device with internet, visit the offline activation page on sugoiaudio.com.
  3. 3. Enter your license key and Machine ID to generate an offline token.
  4. 4. Enter the offline token in the plugin's Settings to activate.

One license per user, tied to your machine.

12. FAQ

What is “mud”? +

Mud refers to excessive low-mid frequency buildup in the 100–700 Hz range. It makes mixes sound boxy, cloudy, and congested — as if you are listening through a cardboard box. Mud is caused by room resonance, proximity effect, overlapping instruments in the same frequency range, or poor microphone placement.

Does Kiyome add latency? +

Minimal. Kiyome uses minimum-phase EQ processing, which adds negligible latency. The AI neural network runs in a background thread and does not add to the audio processing latency. It is suitable for real-time monitoring and tracking.

Can I use it on a mix bus? +

Yes. For mix bus use, switch to M/S mode and use a gentle Depth setting (20–30%). Select the “Full Mix” instrument model. The “Mastering Touch” preset is a great starting point. M/S mode lets you de-mud the center of the mix without affecting stereo width.

What does Learn actually do? +

Learn mode analyzes the spectral energy distribution of your audio over 5 seconds. It identifies which frequency bands have the most energy buildup and automatically adjusts each band's sensitivity to target those areas more aggressively. Bands where the signal is clean receive lower sensitivity. The result is de-mudding that is tailored to your specific material.

Does the AI run on my GPU? +

No. Kiyome uses a lightweight ONNX neural network that runs entirely on the CPU. No GPU, no internet connection, and no cloud processing required. All AI inference happens locally on your machine in real time.

Can I automate parameters? +

Yes. All parameters (Depth, Speed, Tone, Output, instrument selection, band sensitivities, Analog Color amounts, and processing mode) are fully DAW-automatable via the standard AU/VST3 parameter interface.

What is the difference between Speed and Tone? +

Speed controls the reaction time of the spectral detection (temporal domain) — how quickly the plugin responds to changes in the signal. Tone controls the bandwidth of each EQ band (spectral domain) — how wide or narrow the cuts are. Speed is about time; Tone is about frequency width.

Is it safe for mastering? +

Yes. Use the “Mastering Touch” preset as a starting point. Keep Depth low (15–25%), use M/S mode, and always verify with Delta Listen that you are only removing mud, not musical content. The minimum-phase EQ processing preserves phase coherence.

What happens when the trial expires? +

When the 30-day trial expires, the plugin's audio output is muted. The interface remains functional so you can still browse presets and see the controls, but no audio will pass through until you purchase and activate a license.

13. Troubleshooting

Plugin Not Showing in DAW

  1. 1

    Check Plugin Folders

    Verify that Kiyome.component exists in ~/Library/Audio/Plug-Ins/Components/ and Kiyome.vst3 exists in ~/Library/Audio/Plug-Ins/VST3/.

  2. 2

    Rescan Plugins

    Trigger a plugin rescan in your DAW. In Logic Pro, restart the application. In Ableton, go to Preferences > Plug-ins > Rescan.

  3. 3

    Refresh AU Cache

    Open Terminal and run killall -9 AudioComponentRegistrar, then restart your DAW.

  4. 4

    Reinstall

    Download the latest installer from sugoiaudio.com and run it again.

No Processing Happening

  • Check that the Depth knob is above 0%.
  • Verify the Instrument selection matches your source material.
  • Check that the Bypass button is off.
  • Make sure no bands are individually bypassed (right-click nodes to check).
  • Confirm the trial has not expired (expired trial mutes audio output).

UI Blank or White

  • This is a macOS WebKit rendering issue. Restart your DAW.
  • Verify your macOS version is 10.15 or later (Apple menu > About This Mac).
  • Try loading the Standalone app to verify the UI works outside your DAW.
  • Update macOS to the latest version — WebView rendering depends on the system-provided WebKit engine.

Learn Not Working

  • Ensure audio is actively playing through the plugin during the Learn countdown.
  • The signal must be audible — silence or extremely quiet signals will not produce a useful profile.
  • If the button stays in countdown mode, wait for it to complete (5 seconds) or click to stop early.

High CPU Usage

  • Switch to the Full Mix instrument model, which has lower complexity than specialized models.
  • Increase your DAW's audio buffer size (e.g., from 128 to 256 or 512 samples).
  • Close unused plugins and tracks in your session.
  • The ONNX neural network runs on CPU — on older Intel Macs, you may see higher CPU usage than on Apple Silicon.

License Issues

  • Check your internet connection — online activation requires a network connection.
  • Ensure you are entering the license key exactly as provided (no extra spaces).
  • If you are offline, use the offline activation method (Machine ID + offline token).
  • If you have changed machines, contact support to transfer your license.

If the Issue Persists

Please contact our technical support team with the following information:

  • Operating system version (e.g., macOS 14.2)
  • DAW name and version
  • Plugin format used (AU or VST3)
  • Description of the problem and steps to reproduce it
support@sugoiaudio.com →
Back to Kiyome All Products