How to clean and master a track in the browser

Know your console first

A toolbar line opens the page: the drop zone and the remote-link box on one side, the preset list, the Save Current field and the tempo tools on the other. The rack sits under it, and on a full desktop screen the rack is three columns. The deck runs down the left — the waveform and its transport, the comparison button, the True Peak and phase meters, the vector scope with the Mono Check shortcut, the microphone switch and the Studio Report Card. The analysis bays fill the middle column: the twelve-band analyser with Reset to Flat beneath it, the graphic rack, the spectral taming view, and the format picker at the foot. The parameter stack runs down the right, ending in the account bar. Narrow the window and the rack folds to two columns, with the analysis bays and the parameter stack sharing the right-hand side; on a phone, or in a window too short for the compact layout, everything stacks into a single scrolling column. The batch queue sits below the whole rack, and it is where every export trigger lives. Every stage is processed in your own browser, so nothing you load is uploaded anywhere.

Start from a preset, then adjust

A preset is more than a few slider positions. It sets the sliders, the toggles, the whole twelve-band curve and the analyser's target line in one move, which makes it the fastest honest starting point for a genre you do not know well yet.

• Genre starting points: Default Flat Settings, Hip-Hop / Trap / Rap Tuning, EDM / Aggressive Electro Guard, Rock / Heavy Metal Resonance Fix, Pop / Commercial Vocal Clean, Lo-Fi / Ambient Chill Smooth, Country / Warm Acoustic Roots, R&B / Neo-Soul Smooth Balance, Acoustic / Natural Room Honesty, Metal / High-Gain Attack Control, Jazz / Live Room Transparency and Cinematic / Wide Score Depth.
• Special purpose: Warm Analog Tape Glue, and Modern Air and Punch Lift for a finished-sounding top end.
• Utility: Mono Compatibility Check, which folds the mix down and centres it so you hear exactly what a phone speaker will get.

Found something you like? Type a name in the field beside the preset list and press Save Current. Your preset is stored in this browser and comes back in the list after a refresh; it is local to the machine you saved it on, not tied to an account.

Read the analyser against its target line

The twelve columns are the finished chain output, grouped into studio bands from 30 Hz to 16 kHz and weighted so a balanced mix reads as a balanced set of columns rather than a wall of low end. The scale slides with the material: the strongest column is held just under the 48 dB ceiling, so a quiet demo and a hot master both fill the panel, and 0 dB is the silent floor.

The dashed line across the columns is the target tilt for the genre you picked. Its angle runs from 0.5 dB per octave for jazz and acoustic material up to 3.0 for lo-fi and ambient — roughly -3.5 to -6 dB per octave in absolute terms, with 1.5, the balanced modern master, in the middle of the range. A preset you saved yourself falls back to that balanced figure. If the treble end of your trace sits above the line, the top of the mix is brighter or harsher than the genre's own balance; below it, the top is rolled off.

Use the spectral view to confirm the fix

The Spectral Taming View scrolls your spectrum over time, low frequencies at the bottom and high frequencies at the top. Red bands are harsh energy still passing through untouched; blue bands are frequencies the chain has pulled down. Raise the fizz cut, engage the de-esser, or drop the 10 kHz and 16 kHz faders, and watch the red turn blue — it is the visual receipt for the move you just made.

Everything the deck carries

The steps that follow name these controls as they go. This is the full inventory, grouped by what each control does.

The corrective pass, in order

Six stages, run in this order. Each one assumes the one before it is done, because that is the order the problems sit in: cleaning the bottom makes the tone decisions easier, the tone decisions make the loudness decision easier, and the loudness decision is what the export has to carry.

Step 1 — Cut the rumble you cannot hear

Energy below roughly 40 Hz is mostly inaudible on ordinary playback systems, but it still consumes headroom and can push a limiter into pumping. The 40Hz Sub-Steep Guard cuts steeply in that region and buys back dynamic range for the parts of the spectrum that carry the music. The Sub-Bass Dynamic Filter goes a step further, sweeping between 20 Hz and 80 Hz, so a thin mix can pass intact while a muddy one is cleaned up. Both are a single setting for the whole rack, so a plain batch shares whatever you dial in — the one exception is a stem mix, where a row categorized Bass carries the guard on its own path. Listen on headphones while you move the sweep: too high and a bass-heavy track starts to feel weightless.

Step 2 — Shape the spectrum with the graphic rack

Sibilance — the "s" and "sh" sounds in a vocal — lives around 6 to 8 kHz, and a narrow cut of a few decibels is usually enough to stop it stabbing. The Adaptive High De-Esser is exactly that: a 6.5 kHz peaking cut of about 5 dB. Generated vocals often carry a different problem, a glassy, robotic shimmer around 3 kHz, where the AI Fizz Cut takes the edge off without dulling the voice. Both moves are subtractive, and both are easy to overdo — if a voice starts to sound lisping or lifeless, you have gone too far.

The rack beneath the analyser is where the fine shaping happens: twelve bands from 30 Hz to 16 kHz, shelves at each end and ten bells between them, running plus or minus 12 dB. The 3 kHz fader and the AI Fizz Cut are the same band — move either and both follow — which is why the preset curves put their deepest cut there. Work subtractively first: pull the offending band down rather than lifting everything around it, and press Reset to Flat when you want to start again from a clean curve.

Step 3 — Widen the image without losing mono

Stereo width is a mid/side decision: how much of the side signal rides alongside the centre. Widening the sides makes a mix feel bigger, but it also thins the centre image and can push a track out of phase, which is why the Stereo Width Multiplier is level-matched to the track's own side energy — the perceived loudness stays where it was as you turn the control. It runs from 0 to 2x, and 1.0x is untouched. Watch the Stereo Phase Correlation meter as you go: stay in positive territory, then fold the mix down with the Mono Mix Checker — or the Mono Check shortcut pinned to the vector scope — because a phone speaker, a club rig and a smart speaker all reproduce something close to mono.

Step 4 — Pick a loudness target

Streaming services normalise playback, so a track mastered louder than the platform's reference is simply turned down — and loses the dynamics it spent to get there. Around -14 LUFS suits streaming delivery, while -9 LUFS suits club and DJ material that will be played loud in a room without normalisation. The Automated Target Loudness Optimizer does that work for you: the streaming profile compresses gently and leaves about a decibel of headroom, and the club profile compresses harder and leaves almost none. While a profile is active it takes over the compressor and the trim stage; choose Disabled / Manual Calibration to get them back and set Smooth Program Dynamics and the Headroom Trim Target yourself.

Whichever route you take, keep the True Peak Loudness meter comfortably under 0 dBFS and leave a decibel or two of room below it. The meter reads the sample values, and the inter-sample peaks that sit between them can still clip a lossy encoder, so headroom on the way out is not timidity, it is insurance.

Step 5 — Shape the ends of the track

Track Envelope Transitions sets a Fade-In Duration and a Fade-Out Duration, 0 to 10 seconds each, as gain automation on a dedicated stage at the end of the chain. Both windows are measured from the track's own timeline rather than from the moment you pressed play, so starting or scrubbing into the middle of a fade picks the curve up at the level that position sits at instead of jumping to the top of the ramp, and a looped preview opens and closes on its own fades every pass. The curve is rendered into the 16-bit WAV, the 24-bit WAV and the MP3 exactly as it is set, so what you audition is what downloads. In a batch, one envelope applies to every track, which is usually what an album wants.

Step 6 — Export the right format

Export 16-Bit Linear WAV at the session sample rate for distribution and for anything a streaming platform will ingest. Choose 24-Bit Studio WAV when the file is going into a DAW for further work, because the extra precision survives later gain changes. Use the 320 kbps High-Quality MP3 for references, rough shares and anything sent through a messaging app. Type the title, artist and track number into each queue row before you export: they are tagged into the 24-bit WAV and the MP3 (the 16-bit WAV carries the audio alone), and a numbered, tagged batch saves a tedious afternoon when it comes time to upload an album.

Advanced workflow: preparing and cleaning AI stems for your DAW

A final, fully mixed track can be cleaned in a single pass through Audio Cleaner Pro, but the cleanest result comes from working on the separate stems (the individual tracks of a mix — Vocals, Drums, Bass, Synths) before importing them into a Digital Audio Workstation.

When you process a pre-mixed track, your filters are forced to treat the vocals, drums and sub-bass simultaneously. By sanitizing your stems individually through the Studio Rack, you can target the precise frequency flaws of each element without affecting the rest of the arrangement.

🎤 1. Sanitizing AI vocal stems

Generative vocal models are often plagued by a glassy, sharp high-frequency resonance. Left in place, a DAW compressor will squeeze it further forward and make the harshness harder to undo later.

• The protocol: Drop your vocal stem into the drop zone. Move the AI Fizz Cut slider down to carve out the signature 3 kHz metallic harshness, and toggle the Adaptive High De-Esser to clamp piercing sibilance. Keep the Stereo Width Multiplier at a tight 1.0x to hold absolute centre focus for the lead performance.

🥁 2. Carving headroom on kick and bass stems

AI music engines love to dump heavy, chaotic sub-bass rumble into low-end tracks. You cannot hear this rumble on basic headphones, but it eats a massive amount of your dynamic headroom, causing your DAW's master limiters to choke and pump uncomfortably.

• The protocol: Load your bass and drum stems as one batch and engage the 40Hz Sub-Steep Guard. This cuts steeply below 40 Hz — energy most playback systems cannot reproduce, but a compressor still reacts to — so your DAW compressors focus on the punch of the kick drum and the melody of the bassline. In a plain per-track batch the guard is one global setting shared across the run, so two stems needing different guards go out as two passes — or categorize the low-end row Bass, which engages the guard on that row's own path while the rest of the set is summed around it.

🎹 3. Maximizing width on melodic and synth stems

If your backing instruments, pads or sample loops feel thin, flat and pushed into the centre of the mix, the mid/side matrix will build a wider soundstage for them.

• The protocol: Drop your synth or ambient stems into the deck and advance the Stereo Width Multiplier toward 1.5x. Because the matrix is level-compensated against the stem's own side energy, the part expands out to the sides of the stereo field with zero volume drift. Monitor the Stereo Phase Correlation bar to keep the block smoothly in the positive blue zone, and check the mono fold-down — that is what keeps the stem safe on phone speakers.

🎚️ Categorizing a queue: the summed stem mix

The queue knows what each file is. Every row carries a Stem Category selector beside its file name — Unassigned, Drums, Bass or Lead Vocal — and the category is what routes that row's own path. It never changes your file: it chooses the corrections the row carries into the sum. Drums and Bass pass through the brickwall pair, two compressors clamped hard in series, which flattens transients and holds the low end steady; Bass also engages the 40Hz Sub-Steep Guard, so its lowest octave arrives with its headroom intact. Lead Vocal rides the 3kHz Fizz Cut — the console's own cut on that band plus a rider of its own — with a de-esser held on for whatever the cut leaves behind.

From two categorized rows upward the summing engine is armed. The deck stops auditioning one file and plays the whole set at once, each stem through its own path, on the mix's own timeline: the longest stem decides where the mix ends. The batch button becomes Export Stem Mix and writes a single summed, mastered file, balanced by an equal-power downmix so the parts arrive at the level they were heard at. One summed file is one export credit, and every row keeps its own Export trigger for writing that one stem on its own. Leave a row Unassigned and it stays out of the sum.

📦 The multitrack export workflow

Once your settings are dialled in across your catalog, select 24-Bit Studio WAV (HD Precision) as your compilation format and press Export. Drop the freshly cleaned stems straight into timelines like FL Studio, Ableton Live, Logic Pro or Pro Tools. Because the core cleanup is already staged, your high-end DAW plugins can do what they do best: focus purely on creative mixing and creative depth.

The pass carries everything the deck is set to — the graphic curve, the de-esser, the width, the loudness profile or manual dynamics, the headroom trim and the track envelope. It leaves the listening-only tools behind: Studio Space Blend, Monitor Playback Volume and Live Microphone Auditing never reach the rendered file.

Visual telemetry: how to read the Stereo Field Vector Scope (goniometer)

The Stereo Field Vector Scope, or goniometer, is an advanced visual monitoring device that plots your left and right audio channels on a geometric grid. While standard decibel meters only tell you how loud a signal is, the vector scope gives you an instantaneous blueprint of your track's stereo width and structural phase compatibility.

When a song plays, the scope draws a rapidly dancing cloud of lines. Learning to read those shapes is your insurance policy against mixes that fall apart on streaming platforms.

📐 Decoding the visual vectors

• The laser-tight vertical beam (center core):A dominant vertical line slicing straight up the middle of the crosshair axis represents your mono core — the structural foundation of the song. It proves that crucial components like your sub-bass, kick drum and lead vocal are locked dead-centre, which is what keeps the core rhythm loud, punchy and safe from cancellation on a mono phone, a small smart speaker or a club system.
• The circular scribbles (stereo side artifacts):Drag a raw AI-generated track into the deck and you will often see a tight vertical beam surrounded by a frantic, messy circular pattern. That is your side channel information: the content sitting off-centre that disagrees between the two channels, which is what reads as a chaotic soundstage rather than a wide one.
• The flat horizontal oval (the phase clash danger zone):If the cloud flattens horizontally or stretches wide across the left and right borders, this is a critical red flag. It indicates severe phase cancellation between your left and right channels. Sounds sitting in that outer zone will cancel each other out and disappear from the mix the moment your audio drops down to a mono playback system.

🎛️ Reshaping the trace

The three moves that turn a messy scribble into a clean, commercial shape are the ones the steps above already cover. The 40Hz Sub-Steep Guard drops the wild tracking bloat out of the lower quadrant and focuses the centre beam; the AI Fizz Cut pulls the frantic outer scribbles inward; and the Stereo Width Multiplier is what shapes the cloud itself — at 0 it collapses to a single vertical line, and back out around 1.1x or 1.2x it opens into a controlled oval that stays safe for public distribution. The Reset to Flat and Mono Check shortcuts are there for exactly this kind of comparison: flatten the rack or fold the mix down without stopping the transport.

Live room tools: microphone auditing and ducked studio ambience

To move from file-based track editing into live content production, the console carries a live microphone auditing engine and an automated, side-chained studio ambience branch. Both run in the browser and both are monitoring tools, which is worth saying plainly: they change what you hear, never what you export.

🎙️ 1. Real-time live microphone auditing

When you are prepping a livestream, recording a podcast or tracking vocals, flying blind on your room acoustics leads to bad results: computer fan hum, air-conditioning frequencies and reflections from untreated walls can mask the performance without you noticing.

• The workflow: Engaging Live Microphone Auditing asks the browser for microphone permission and streams the capsule input straight into the processing chain. The device is released again the moment you switch it off, so no tab holds the microphone open behind your back.
• The telemetry: Your live voice immediately drives the True Peak meter, the phase correlation bar, the vector scope, the twelve-band analyser and the spectral view. You can pinpoint reflections, adjust your seating distance from the mic, dial in the fizz cut for harsh room sibilance, and hear your room's own balance before it reaches your broadcast software.
• Feedback protection: While the microphone is live, the speaker output is held at zero by default — a microphone through a processing chain with the speakers open is a feedback loop. The instruments all read the chain before that stage, so they stay useful while the output is muted. Deliberately raising the Monitor Playback Volume is taken as a decision to listen on headphones, and it lifts that mute.

🌌 2. Auto-ducked studio ambience (the anti-dryness mask)

A common side-effect of aggressive filtering, high-Q notches and steep low-end guards is that the result sounds too dry, clinical and unnaturally isolated — in engineering terms, room suffocation. The voice lacks life because the natural space around it has been stripped away.

• The dynamics engine: The Studio Space Blend drops in a parallel branch mimicking a tight, well-treated voiceover booth — a short, dark tail, band-limited so it reads as closeness around the source rather than a wash behind it. A constant reverb would muffle speech and ruin your presence, so this one is not constant.
• The side-chain solution: An envelope follower tracks the amplitude of whatever is playing. The moment you speak or a track plays, the engine ducks the ambience by -8 dB, keeping dialogue front, clear and punchy. The moment you take a breath, pause or hit silence, it ramps back up over a roughly 300 ms release, so the soft room air blooms out and masks the edits, breath and gating artifacts that heavy cleaning leaves behind.
• Listening only: The ambience branch is never part of the offline render, and at 0 percent it is unplugged from the graph entirely. You may see the vector scope open up as you raise it; your exported file will not carry it.

🎛️ Mastering the live console mix

Use the Studio Space Blend slider to balance your environment. For voiceovers and clean narration, a low blend of 5 to 15 percent provides just enough air to keep the track sounding organic. For ambient music or cinematic vocals, push the blend up to 30 percent and watch the vector scope expand from a tight mono line into a spacious, controlled field. The Monitor Playback Volume (Listening Only) fader sits in the same stack and behaves the same way: it is your speaker level, it is outside the chain, and it can never change a rendered file.

Working in batches

The batch queue takes as many files as you drop on it, decodes them one after another and exports each as its own file. Every row carries its own title, artist and track number, tagged into the 24-bit WAV and the MP3 — a numbered, tagged batch saves a tedious afternoon at upload time.

• One setting, every track: the preset, the graphic curve, the loudness profile and the track envelope apply to the whole batch. That is the point of it — an EP or album stays consistent from track to track instead of being shaped five slightly different ways.
• Auditioning a row: click any queue row to load that track into the deck and listen. Selecting a row never changes what the batch exports; the export walks the queue from the top.
• Loading from a link: the Stream from remote Web URL box fetches a file over http or https without downloading it first. The host has to permit cross-origin reads, or the browser blocks the request.
• Categorizing rows into a stem mix: every row carries a Stem Category — Unassigned, Drums, Bass or Lead Vocal. From two categorized rows upward the queue stops working track by track: the deck auditions the set as one performance through the path each category asks for, and the batch button becomes Export Stem Mix, writing the set as a single summed and mastered file for one export credit. Every row keeps its own Export trigger for a single stem, the header line says which of the two the button will do, and a row left Unassigned stays out of the sum.
• What the whole run will cost: the line in the queue header prices the batch before you commit to it — one credit per track on a credit balance, one of the three daily free exports on the free tier with the slots left named, and Pro members reading that the run uses no credits at all. A track that has not finished decoding is skipped rather than charged, so the figure in the header is exactly what the run spends.
• Credits, track by track: each exported track spends one export credit, or one of the three free exports every account gets each day; when the day's allowance and the balance are both gone, the batch stops and the plan dialog opens with the two ways past it. An unlimited plan exports the whole queue without counting.
• Taking tracks back out: every row carries a ✕ that drops that one track, and Clear Queue beside the export button empties the list in one move. The stored copy in your browser goes with the track, so a queue you have emptied does not come back on your next visit — but the audio on your own disk, your console settings and your presets are all left exactly as they were. Both controls are held shut while an export is running, so a render is never pulled out from under it.
• Your work survives the trip: signing in, or leaving for a checkout, takes you off the page. The queue, the titles you typed and every control position are kept in your own browser and put back when you return, so the console looks the way you left it. Nothing is uploaded to do that.

📸 The Studio Report Card

The Studio Report Card stitches the twelve-band analyser, the stereo vector scope and the True Peak meter into a single 1200 by 400 PNG and downloads it, stamped with the active genre curve. It is read-only — nothing about the chain, the queue or your account changes — which makes it a clean way to show a client where a track sits before and after the pass.

Frequently asked questions

Are my audio files uploaded to a server?

No. Audio Cleaner Pro decodes, processes and renders every file locally in the browser using the Web Audio API. Your track never leaves your machine, which also means processing speed depends on your own hardware rather than on a connection.

Is a browser tool a replacement for a mastering engineer?

Not for a finished commercial release that needs taste, room acoustics and a second pair of ears. It suits a corrective pass: clearing rumble, controlling sibilance, fixing stereo problems and landing a track on a sensible loudness target before it goes out or goes to a mastering engineer.

What does the free part include?

Loading, previewing, the presets, the graphic rack, every meter and the Studio Report Card are free and unlimited, and a signed-in account comes with three free exports a day. Past that, exporting is what you pay for: a $3 pack of twenty full-quality exports that never expire, or an unlimited monthly plan for people releasing regularly.

Can I process a whole album in one pass?

Yes. The batch queue takes multiple files, decodes them one after another and exports each with the title, artist and track number its row carries, tagged into the 24-bit WAV and the MP3. Settings stay the same across the batch - the preset, the graphic curve, the loudness profile and the fades - which is what keeps an EP or album consistent from track to track. The queue header prices the whole run in export credits before you start it, and tracks can be taken back out with the ✕ on their row or emptied out entirely with Clear Queue.

Can I export several stems as one summed mix?

Yes. Give each queue row a Stem Category - Drums, Bass or Lead Vocal - and from two categorized rows upward the summing engine is armed: the deck auditions them together as one performance, each stem through the corrections its category asks for, and the batch button becomes Export Stem Mix, writing a single summed and mastered file. A row left Unassigned stays out of the sum, and every row keeps its own trigger for exporting one stem on its own. One summed file is one export credit.

Will my fades be written into the exported file?

Yes. Fade-in and fade-out are baked into the 16-bit WAV, the 24-bit WAV and the MP3 exactly as they are set. Both windows are measured from the track's own timeline, so starting or scrubbing into the middle of a fade picks the curve up at the level that position sits at.

Do the ambience or the microphone tools change my exported file?

No. Studio Space Blend, Monitor Playback Volume and Live Microphone Auditing are all listening-only. The export carries your gain staging, the sub guard, the sub-bass filter, the graphic curve, the de-esser, the stereo width, the mono state, the dynamics or loudness profile, the headroom trim and the fades.

Will I lose my queue if I sign in or pay?

No. The queue, the titles you typed and every control position are kept in your own browser and put back when you return from a sign-in or a checkout, so the tool looks the way you left it. Nothing is uploaded.