StemsAI Mixer v1.0 Beta – User Guide & Workflow

← Home  |  Open Mixer →

This guide shows you how to take raw AI stems from Suno, Udio and others, clean up harshness and wide bass, add some human movement, and export a mix – all in your browser.

About this version: StemsAI Mixer v1.0 is a Beta built on the Web Audio API. It’s tuned for short projects and AI stems rather than huge multi-track sessions. Expect a few rough edges.

1. Import & Naming 2. Track Strips 3. AI Stem Cleaners 4. Humanizer Panel (Beta) 5. Sample Rack & Clip Markers 6. Sidechain Ducking 7. Master, Normalize & Reverb 8. Record vs Bounce 9. Limits & Best Use Cases 10. Suggested Workflow

1 Import & Rename Your Stems

Open StemsAI Mixer and either:

• Click the file picker in the top bar, or
• Drag & drop your stems (WAV or MP3) anywhere in the window.

Each file creates a track strip in the lower Tracks area.

Why naming matters (Roles)

StemsAI uses your track names and their Role selection to drive AI helpers and mono-bass logic. Generic names like audio_1.wav give it nothing to work with.

2 Track Strips – Core Controls

Each stem gets a strip with everything you need for a basic AI mix cleanup.

Top row: Name, Role, Mute/Solo

• Name – click to rename the track.
• Role – tells the mixer what this stem is (Kick, Bass, Vox, Pad, etc.).
• Mute – silence just this track.
• Solo – hear only this track (or group of solos).
• Reset – reset this strip to safe defaults.

Level, Pan & Width

• Gain – track volume.
• Pan – move the sound left/right.
• Width – Mid/Side stereo width:
  • 0.0 = mono (centre only).
  • 1.0 = normal stereo.
  • 2.0 = extra wide (good for pads/FX, not good for bass).
• Pitch (st) – semitone pitch adjust, useful for tiny tuning fixes and layering.

Filters & EQ

• HP – high-pass filter; clears low rumble on vocals/pads.
• LP – low-pass filter; softens harsh highs.
• Low / Mid / High EQ (dB) – 3-band tone control.

Delay, Chorus & Phaser

• Delay – time/feedback/mix, plus tempo sync options.
• Chorus – subtle movement; best used in small amounts.
• Phaser – swirling motion; nice on pads, easy to overdo.

Reverb Send & Duck Amount

• Rev Send – how much of this track hits the global reverb.
• Duck – how strongly this track ducks under your chosen Duck key (sidechain).

3 AI Stem Cleaners

The AI Stem Cleaners panel is there to fix common AI-artifact problems fast.

De-Haze LP

• Choose a cutoff with the De-Haze LP slider (e.g. 12 kHz).
• Click AI De-Haze (Non-Bass).
• The app:
  • Skips low-end roles (Kick/Bass/Sub/808).
  • Applies a gentle LP filter to bright tracks to reduce harshness.

Bass Mono-Maker

• Click Bass Mono-Maker.
• Tracks tagged as Bass / Sub / 808 are:
  • Re-patched into a Mid/Side matrix so their low frequencies are centered.
  • Kept mono-compatible so clubs and phones don’t lose the low-end.

AI De-Verb (Vox / Lead)

• Select your vocal or lead tracks (Role: Vox/Lead).
• Click AI De-Verb (Vox / Lead).
• The app tries to pull back excessive reverb “wash” and tame tails, without fully drying out the track.

4 Humanizer One-Clicks (Beta)

The Humanizer panel adds subtle (and sometimes not-so-subtle) movement so AI stems feel less robotic. These are designed around trance/EDM-ish material but can work on other genres with care.

For best results, turn on Write automation before you click these, so the moves are recorded into your project and picked up by the Record path.

What each button does

• Micro Pitch Drift – slow ±cents pitch wobble on musical roles (pads, leads, vox) to fake tape drift.
• Human Duck Wobble – gently varies ducking depth so sidechain doesn’t feel “machine perfect”.
• Breathing Chorus – low-rate, low-depth chorus with its parameters moving over time.
• Vocal Air Ride – a moving high-shelf on vocal/lead roles for subtle air and presence changes.
• Delay Throw – pushes a one-shot burst of delay on suitable tracks (throw effect).
• ⚡ FULL PRO HUMANIZE – fires a heavier chain of the above, with layered timing changes.
• Reset Humanize – clears the Humanizer moves back to a safe-neutral state.

5 One-Shot Sample Rack & Clip Markers

One-Shot Sample Rack

The Sample Rack lets you sprinkle in extra FX, shouts or stabs without leaving the browser.

• Drop short samples (vocal shouts, impacts, risers) into the rack area.
• Trigger them live while recording, or automate them into the timeline.
• They route through the same master chain, so they match your mix space.

Clip Markers

Whenever the master output hits 0dB and clips, StemsAI logs a clip marker with a timestamp.

• Use the Clip Markers panel to find where things are getting too hot.
• Back off track gain or master gain, or tweak compression/limiter as needed.
• The goal is to keep red clips rare or non-existent before export.

6 Sidechain Ducking (Duck Key)

StemsAI includes a simple sidechain ducking system for that classic “AI stems but actually pumpy” feel.

1. In the header, set BPM to match your track.
2. Load and rename your stems, setting roles where possible.
3. Choose a Duck key – usually your Kick or Drum Bus – from the dropdown.
4. On other tracks (Bass, Pads, Leads), use the Duck slider to set how hard they duck.

Internally, the app measures the loudness of the Duck key and applies a smooth envelope to other tracks. You get clean pump, without horrible clicks.

7 Master, Normalize & Global Reverb

Master Controls

• Clip Guard – auto-reduces master gain if clipping is detected.
• Gain – overall level into the master chain.
• Comp Thr / Ratio – glue compression for the whole mix.
• Limiter – soft-clip limiter to catch peaks.
• Meter – visual peak meter; aim to live mostly in green/amber.

Normalize to -12 LUFS (Experimental)

The Normalize to -12 LUFS button is a small helper that tries to push your mix toward a streaming-friendly loudness target.

• It estimates loudness and adjusts master gain accordingly.
• It’s not a full mastering chain – think of it as a first pass.
• You can still tweak the master after hitting it.

Global Reverb

• Return – how loud the shared reverb is in the mix.
• Color – low-pass filter on reverb (higher = brighter).
• Length – decay time.
• Rebuild IR – regenerates the reverb “shape” with the current settings.

8 Export – Record vs Bounce

🎧 Record (WEBM – true “what you heard”)

• Click ● Record, then press Play.
• Perform your moves: tweak knobs, trigger samples, click Humanizer buttons.
• Click ■ Stop Rec. A Download mix.webm link appears.

The WEBM file is a direct capture of the master output. It includes:

• Sidechain ducking behaviour.
• All Humanizer timing moves.
• Sample rack hits, automation and master processing.

If you want exactly what you heard, use Record.

📀 Bounce to WAV (static mixdown)

• Click Bounce to WAV (or Bounce Loop if you set In/Out points).
• StemsAI rebuilds a similar chain offline and renders mix.wav.

The WAV includes:

• Per-track EQ, filters and width.
• Delay, chorus, global reverb.
• Master compression and limiting.

But in this Beta, Bounce is a snapshot. It does not currently replay:

• Real-time sidechain ducking envelopes.
• Humanizer timing logic (Micro Drift, Wobble, Delay Throw, etc.).
• The full dynamic LFO behaviour used in live playback.

9 Limits & Best Use Cases (v1.0 Beta)

StemsAI Mixer v1.0 Beta is intentionally small and focused. It’s not a full DAW replacement.

Recommended limits

• Stems: Up to ~6 stems per project is the “happy path”.
• Length: Around 4 minutes per project/loop is what it’s tuned for.
• Browser: Desktop Chromium-based (Chrome/Edge/Brave). Not tested for long sessions on mobile.
• Material: Works best on unmastered AI stems, not on full squashed masters.

Great fits

• Cleaning up Suno/Udio tracks before upload.
• Adding light movement to trance/EDM style stems.
• Quick DJ edit-style tweaks: de-harsh, mono bass, small air & width.

Not ideal for

• Huge 20–30 stem projects, long multi-hour sets.
• Heavy, already-mastered rock/metal where everything is pinned to 0dB.
• Precision album mastering or stem mastering – use a full DAW for that.

10 Suggested 5-Minute Workflow

1. Import & rename: Drag in your stems, rename them and set sensible Roles (Kick, Bass, Vox, Pad, etc.).
2. Clean extremes: Run AI De-Haze on bright stuff and Bass Mono-Maker on the low-end. Use AI De-Verb on smeared vocals.
3. Add gentle movement: Try Micro Pitch Drift + Breathing Chorus on pads/leads, and maybe a touch of Vocal Air Ride on vox. Leave FULL PRO HUMANIZE off unless you know the track likes it.
4. Balance the mix:
   • Kick in the middle; bass mostly mono.
   • Pads/FX wide; hats somewhere in between.
   • Use HP/LP and 3-band EQ to carve space.
   • Add a bit of global reverb for glue.
5. Export:
   • Record → mix.webm if your performance + Humanizer movement is important.
   • Bounce to WAV if you want a static, clean file to master further or upload.