MasterForge Pro Master
User Guide

01 Quick Start

Get from raw audio to mastered output in five steps. No experience needed.

What is mastering?

Mastering is the final polish before release. It's not mixing (balancing individual tracks) — it's shaping the overall sound: loudness, frequency balance, stereo image, and dynamics. Think of it like color grading a film after editing.

Your first master

1. Drop your audio file onto the Pro Master screen (or click to browse). Supports WAV, MP3, FLAC, OGG, and AAC — up to 32-bit float, any sample rate.
2. Choose a preset from the dropdown at the top. There are 13 factory presets covering Pop, Rock, EDM, Hip-Hop, Jazz, Classical, and more. Gentle Master is a great starting point.
3. Press Play and listen. You're hearing the processed version (B). Press A to hear the original for comparison, B to switch back.
4. Adjust if needed. ToneMap (Chapter 4) lets you dial in warmth, space, width, and edge with four simple sliders. Or tweak individual modules.
5. Export. Click Export, choose 16-bit WAV (streaming) or 24-bit WAV (archival), and download your mastered file.

A/B comparison tips

When comparing processed (B) vs. original (A), the processed version is usually louder. Louder sounds "better" even when it isn't. Try to judge quality not volume — listen for clarity, bass tightness, and detail rather than raw level. The bypass path is latency-compensated, so timing stays perfectly aligned.

The floating player

Play, stop, and seek follow you in a small player pinned to the bottom right of the screen, so you can start and stop no matter how far you have scrolled into the modules. If you close it with the ×, click Player on the waveform toolbar to bring it back. Your choice is remembered next time.

02 Signal Chain

Your audio flows through 12 processing stages in a specific order. Understanding this order helps you make better decisions — each stage builds on the previous one.

Color coding

Throughout this guide, processing stages are color-coded by category:

I/O Input/Output gain Frequency EQ, HPF AI Tools Artifact Supp, De-Esser Stereo M/S, Width Enhancement Sub-Bass, Exciter Dynamics Comp, Limiter, Sat

A/B bypass path

When you press A (bypass), audio flows through a separate delay line that exactly matches the processing latency of the full chain. This means A and B are time-aligned — you're hearing a true comparison, not a delayed echo. Maximum latency in HQ mode is approximately 149ms (linear-phase EQ + AI Artifact Suppressor + Spectral Suppressor), all automatically compensated.

03 Parametric EQ

The EQ is your most powerful frequency-shaping tool. Double-click anywhere on the EQ display to create a band, then shape it with your mouse.

Mouse interactions

You can have up to 31 bands simultaneously.

Filter types

Understanding Q

Q controls how wide or narrow an EQ band is. Think of it as a spotlight: low Q is a floodlight (affects a wide range), high Q is a laser (pinpoints one frequency).

EQ modes

Min-phase (default)

Zero latency, uses the browser's native BiquadFilter (C++ code, very efficient). Introduces minor phase shift at boost/cut frequencies — this is normal and usually inaudible. Best for real-time monitoring.

Linear-phase (HQ mode)

FFT-based processing via WebAssembly. Adds 64ms of latency (3072 samples at 48kHz), but introduces zero phase shift — preserving transient shape perfectly. Better for final mastering. The latency is automatically compensated in A/B bypass comparison. Switch via the Quality Mode selector or the EQ panel dropdown.

Match EQ

Match EQ listens to a commercial reference track and steers your master's tonal balance toward it. Load a reference in the Reference strip (Choose file, or drag audio onto it), then click Match EQ. A preview appears comparing your current EQ (gold) against the proposed match (green dashed), so you approve or cancel before anything changes.

It compares average spectra, not moments. The reference is measured against a 10 to 15 second slice of your master (your loop region if you set one, otherwise the loudest section), rendered through your full chain. The overall level difference is removed first, so Match EQ chases tone, never loudness.

04 ToneMap

ToneMap is a meta-controller — it doesn't add its own processing. Instead, it physically moves 26 sliders across other panels (EQ, Compressor, Stereo, Saturation, De-Esser) simultaneously. One-minute mastering with four intuitive axes.

The four axes

Each axis ranges from -100 to +100. Double-click any axis to reset it to zero.

Safety zones

0–60: Safe Musical adjustments, always sounds good. 60–80: Experimental Stronger effects — check carefully. 80–100: Extreme May cause mud, harshness, or phase issues.

The UI shows color feedback in real-time: green → orange → red as you push further.

What ToneMap drives

Behind the scenes, ToneMap controls 26 parameters across multiple panels:

Stereo: width, mid gain, side gain, bass mono frequency, side high-shelf. M/S EQ: 5 sub-bands (mid lo/mid/hi, side lo/mid). Saturation: drive amount. Compressor: per-band attack, release, and ratio (4 bands × 3 = 12 parameters). AI tools: artifact fog, shimmer threshold, de-esser range.

ToneMap also creates its own EQ bands: a low shelf at 200 Hz (driven by Warmth) and a high shelf at 8 kHz (driven by Warmth + Edge).

ToneMap presets

Eight quick-start profiles: Cozy Evening, Crisp Pop, Club Ready, Podcast Warm, Headphone Intimate, Speaker Room, Vintage Vinyl, Modern Bright. Select one as a starting point, then fine-tune the axes.

05 Dynamics

The dynamics section handles three processors in a single AudioWorklet: multiband compressor, saturation, and limiter. The limiter is always last in the chain — everything else feeds into it.

5.1 Multiband Compressor

Splits your audio into 4 frequency bands using LR4 crossovers (default: ~200 / ~2000 / ~8000 Hz), then compresses each band independently. This lets you tighten the bass without affecting vocal dynamics, or control bright cymbals without dulling the midrange.

Per-band parameters

Auto-Makeup Gain

When enabled, automatically compensates for gain reduction so perceived loudness stays consistent as you adjust compression. Uses a smoothed average of gain reduction (τ=500ms).

Parallel Blend (NY Compression)

Mixes compressed and original signal: 0% = pure dry, 100% = pure compressed. The classic "New York compression" technique uses 30–50% — you get the punch and density of compression while preserving the natural transients of the original.

GR Meters

Per-band gain reduction readout. Yellow = light compression (>0.5 dB). Orange = working hard (>3 dB). If you see constant heavy reduction, consider raising the threshold.

5.2 Limiter

The limiter is a brick-wall ceiling — nothing goes above the level you set. It's the last processor in the dynamics chain (and therefore the last processor before output).

5.3 Saturation

Adds harmonic warmth and character. Three modes offer different coloring: Tube (warm, even harmonics), Tape (soft compression + harmonics), and Clip (hard-edged, aggressive).

Oversampling prevents aliasing artifacts: 1× in Lite, 2× in Normal/HQ, 4× in Export.

5.4 Visual Compressor

The Visual Compressor is a friendlier face on the same 4-band compressor from 5.1. Instead of setting threshold, ratio, attack, and release by hand, you watch a live spectrogram and dial a single Amount per band. Where a band is compressing, a translucent red overlay glows over the spectrum, so you see the squeeze as it happens.

Auto-Fix

Auto-Fix analyses the loudest 8 to 30 seconds of your track and proposes a tasteful per-band "glue" starting point. Click Review to see the suggestion band by band, then Apply to load it, or Re-analyze after you change the track. It is a starting point, not a diagnosis: it aims for gentle cohesion rather than waiting for a problem to appear.

06 Stereo & Spatial

Control how your audio sits in the stereo field. All stereo processing uses Mid/Side (M/S) encoding — the center of your mix (vocals, bass, snare) is the Mid channel, and everything pushed to the sides (reverb, wide guitars, ambience) is the Side channel.

6.1 M/S Stereo Controls

6.2 M/S EQ

Shape the frequency balance of mid and side channels independently. Six bands: Mid Lo, Mid Mid, Mid Hi, Side Lo, Side Mid, Side Hi — each ±12 dB. For example, boost Side Hi for "air" and shimmer, or cut Mid Lo to tighten the bass center.

6.3 Multiband Stereo Expander

Three bands (Lo/Mid/Hi, adjustable crossovers at ~250 Hz and ~4 kHz) with per-band width from 0–200%. This lets you keep bass tight and mono while widening the highs — exactly what most masters need.

Correlation Meter

Shows how "compatible" your stereo image is with mono playback. Mono matters because many listening situations (phone speakers, Bluetooth speakers, some PA systems) collapse stereo to mono.

6.4 Visual Stereo (Simple)

Switch the Stereo panel from Advanced to Visual to see your stereo image as a shape instead of sliders. In these graphs frequency runs up the vertical axis (20 Hz at the bottom, 20 kHz at the top) and width runs left and right from a center mono line. A live Mid/Side spectrum sits behind the shape so you can see what you are working on.

Simple mode is a three-band "hourglass". The horizontal reach at each height is that band's width: pinched to the center line is mono, the dashed lines mark neutral (100%), and full reach is maximum width. Drag a band left or right to set its width, drag the dashed crossover lines to move the band splits, and drag the green line to set the bass-mono frequency. It is a skin over the existing Multiband Width, Mid/Side gain, and Bass Mono controls, so anything you do here moves those same sliders.

6.5 Precise Stereo (spectral)

Precise mode replaces the three bands with two continuous curves you draw yourself: a Side curve that sets width at every frequency, and a Mid curve that shapes the tone of the center. Double-click to add a point, drag to move it (both edges move together so the image never leans left or right), and right-click to remove one. Each point shows its exact frequency and gain, up to 15 dB, from 20 Hz to 20 kHz.

Precise runs on a linear-phase spectral engine for clean, phase-accurate width shaping. Because that engine works in frequency blocks it adds about 61 ms of latency, which is compensated automatically in A/B so the comparison stays aligned. Turn it on with the On (HQ) checkbox in the Precise header, and use Clear curves to start over.

07 AI Audio Tools

MasterForge is built for AI-generated music. These tools target specific problems created by neural audio codecs (Suno, Udio, and similar generators).

7.1 De-Esser

Reduces harsh sibilant sounds (s, sh, ch). AI generators often produce exaggerated sibilance because neural codecs struggle with high-frequency transients.

7.2 AI Artifact Suppressor

The flagship AI tool. Targets three specific artifact types created by neural audio generation:

Shimmer

Metallic, ringing artifacts in the 6–14 kHz range caused by neural codec quantization. Sounds like a tinny overlay on cymbals and vocal sibilants. The suppressor detects unstable spectral bins in this range and attenuates them.

Fog

A "foggy" quality in the 400 Hz–2 kHz range caused by diffusion model over-smoothing. Makes vocals sound distant and muffled. The suppressor detects unnaturally flat spectral regions and restores dynamic variation.

Pitch artifacts

Phase deviation artifacts where the AI's pitch tracking fails. Causes subtle wobbling or doubling effects. Detected via phase coherence analysis.

7.3 Vocal Clarity Engine

A meta-controller (like ToneMap) that improves vocal presence with a single "Clarity" slider. Behind the scenes, it drives three parameters:

Mud Cut: lowshelf at 300 Hz, up to -3 dB — clears the low-mid muddiness that AI generators often produce. Presence: peaking boost at 3.5 kHz (Q=0.8, wide and musical), up to +2 dB — brings vocals forward. De-Esser offset: up to -3 dB threshold reduction — automatically tames any harshness the presence boost might reveal.

Auto-enables the de-esser when Clarity reaches 20 or above. Has its own bypass toggle for quick A/B.

7.4 Spectral Resonance Suppressor

STFT-based detection and reduction of ringing frequencies. Not specific to AI audio — useful for any source with resonant peaks. Adds 42.7ms latency (2048 samples). User-controlled bypass.

08 Shaping Tools

Three enhancement processors that add punch, weight, and sparkle. All three feed into the dynamics worklet — the limiter catches any energy they add.

8.1 Transient Shaper

Controls the attack and sustain of sounds independently from their volume.

8.2 Sub-Bass Enhancer

Detects the fundamental pitch of your bass content and synthesizes a sub-harmonic one octave below. Adds physical weight you can feel on a system with a subwoofer.

8.3 Harmonic Exciter

Generates harmonic overtones for brightness and "air" — the subtle shimmer that makes a track feel polished and expensive.

09 Multi-Track Stem Mixer

Load individual stems (vocals, drums, bass, etc.) and mix them before mastering. Each track gets its own processing chain, and the summed output feeds into the full mastering signal chain.

Loading stems

Drop multiple audio files onto the dropzone simultaneously, or use "Add Track" after your first file. MasterForge auto-detects Suno/Udio stem naming: "Song Name - Vocals.wav" gets labeled "Vocals" with green color coding. Auto-detected types: Vocals, Drums, Bass, Piano, Keys, Guitar, Synths, Pads, Strings, Instrumental, Other.

Per-track processing chain

Each track flows through: Source → Volume Gain → Pan → EQ Lo (200 Hz shelf) → EQ Mid (1 kHz peak) → EQ Hi (8 kHz shelf) → Compressor → Mute Gain → Sum Bus. The sum bus connects to the master chain's Input Gain, where the full mastering chain applies to the combined signal.

Track strip controls

Sidechain compression

Select a trigger track (e.g., Drums → duck Bass) to create a sidechain relationship. Parameters: threshold, ratio, attack, release, range. Classic technique for making bass "pump" with the kick drum.

Track Edit View

Click "Edit" on any track strip to open a zoomable waveform editor. Ctrl+Scroll to zoom (1×–64×). Click-drag to select a region, then apply Gain Attenuate (-60 to 0 dB with 5ms crossfade) or Region EQ (3-band, processed offline). Delete silences the selection. Ctrl+Z to undo (20-entry stack).

Bounce to Master

After mixing your stems, click "Bounce" to create a summed stereo buffer that includes all per-track processing. This becomes a single audio buffer for the mastering chain.

10 Metering & Analysis

Meters don't lie — they tell you what's actually happening to your audio, regardless of what your speakers (or ears) might suggest.

10.1 Loudness (LUFS)

LUFS (Loudness Units Full Scale) is the industry standard for measuring perceived loudness. Three measurements are available:

Integrated: overall loudness of the entire track. Short-term: 3-second sliding window — shows loudness trends. Momentary: 400ms window — shows moment-to-moment dynamics.

Streaming targets

10.2 True Peak

Digital audio can have "inter-sample peaks" — moments where the reconstructed analog waveform between two samples exceeds 0 dBFS, even if no individual sample does. True Peak measurement catches these. Target: keep True Peak below -1.0 dBTP for streaming — this prevents clipping after the streaming service converts to lossy codecs (AAC, Ogg Vorbis).

10.3 Analysis

The spectrum analyser shows frequency content in real-time using a 4096-point FFT with 0.8 smoothing. Pre-EQ and post-processing views let you see what your processing is doing. The waveform view supports click-drag to set loop regions for focused listening.

11 Export

Export renders your mastered audio offline with the same processing as preview — no hidden changes. Export uses maximum quality: 4× oversampling (vs 2× in Normal mode) and linear-phase EQ.

Format options

Dithering

When you reduce from 24-bit or 32-bit float to 16-bit, tiny rounding errors occur. Dithering adds carefully shaped noise to mask these errors, resulting in a smoother, more natural sound at low levels.

None: raw truncation (only suitable for 24-bit). POW-R Type 1: gentle noise shaping — transparent and safe. POW-R Type 3: aggressive noise shaping — pushes dither noise to frequencies your ear is least sensitive to. Best choice for 16-bit final masters.

LUFS normalization

Optionally normalize your export to a specific loudness target: -14 LUFS (Spotify/YouTube), -16 LUFS (Apple Music), -23 LUFS (Broadcast), or Off (keep current levels).

Output filename

Downloads as {original_name}_mastered_{bitdepth}bit.wav

12 Surgical Cleanup

Surgical Cleanup is a magnifying glass for one bad frequency. AI-generated audio often hides a single metallic ring or a honky resonance that no broad EQ move can fix without dulling everything around it. This tool lets you find that exact frequency, listen to it in isolation, then punch a narrow, adjustable hole in it and leave the rest of the track untouched.

Opening the dock

Load a track first, then click Surgical on the waveform toolbar. The dock opens below the waveform and the waveform switches into a violet cleanup view. The dock header shows which track you are working on.

Find the frequency (Scan)

The scanner shows a live spectrum from 20 Hz to 20 kHz. Drag left and right across it to sweep the target frequency, shown top right (for example "1.00 kHz"). Two listen modes help you hunt:

Scope and bandwidth

Choose which part of the stereo image to treat, and how wide the cut is:

Apply and stack

When the cut sounds right, choose whether it treats a painted Section of the waveform or the Whole track, then click Apply & Save. The cut is baked in with short 5 ms fades at its edges so there are no clicks, and it appears in the Applied Cuts list with a colored dot for its scope (gold Mid, violet Side, cyan Full). You can stack several cuts to clean up multiple frequencies. Cuts are remembered per track, so switching away and coming back keeps them.

13 Monitor & Translation

A master that sounds great on your headphones can fall apart on a phone speaker or in a car. Monitor mode lets you audition your master through those real-world systems while you work, so you can fix translation problems before you export. It changes only what you hear. The exported file and every meter always reflect the real, untouched master.

Target systems

The Monitor buttons in the header switch between four playback simulations, plus Off:

Correct for your own headphones

For the simulation to be accurate, MasterForge needs to account for the coloring your own headphones add. Click the headphones button next to the Monitor targets and pick your model. MasterForge then corrects your headphones toward neutral, using measured AutoEq curves, so the target you hear is honest. Over 30 models are available across the common families, including Sennheiser HD, and your choice is remembered for next time.

Loudness stays fair

Each target naturally changes the level, and louder always seems better. Monitor mode matches loudness automatically using a slow average, so switching targets compares tone, not volume. It follows the track's average level rather than reacting to the music, so it never pumps or breathes.

14 Batch Mastering

Once you have a chain you like, Batch Mastering applies it to a whole set of tracks at once. It captures your current mastering chain and your current export settings, runs every file through them offline, and hands you the results as a single ZIP. Nothing is re-tuned per file, so a batch is only as good as the chain you set up first.

Getting in

The Batch master button sits under Export and lights up once a track is loaded. Batch mastering is a signed-in Pro member feature, so if you are not signed in the dialog offers a login link.

Building the queue

The track you already have loaded is added as the first row, marked Preview (the file you tuned the chain on). Drag more files onto the dialog or use Add files. Supported formats include WAV, MP3, FLAC, OGG, AAC, M4A, and AIFF. A batch holds up to 10 tracks, the preview included.

Running it

The settings line shows exactly what will be printed: output format, sample rate, and LUFS target, all read from your main Export controls. Click Master all to process the queue one file at a time. Each row shows its own progress (waiting, processing, done, or error) and an overall bar tracks the set. Cancel stops after the file currently rendering.

◆ Quality Modes

Auto-detection runs after 10 AudioWorklet benchmark samples. You can override manually at any time.

◆ Factory Presets

◆ Keyboard Shortcuts