A music tracker, sample workshop, effects mixer and audio-reactive visual performance environment in one browser application.
Maxxlorr Tracker (MLT) makes music through patterns: vertically arranged rows of notes, instruments, volumes and compact commands. Patterns form a song in the sequencer. Instead of hiding musical instructions behind a timeline, the tracker keeps them visible and editable, down to individual ticks.
It brings this classic method together with multi-note tracks, graphical command editing, stereo sampling, FM instruments, native DSP racks, MIDI control and the Visual Lab. Use it to compose a piece from scratch, study and rework a tracker module, design an instrument, build a live performance, or turn a composition into sound and moving images.
MLT 5.04: Pattern view: a generated six-track demonstration, with a three-note chord track, a two-command track, instrument colours and graphical effect controls.
Who it is for
Tracker musicians who want precise, compact composition and modern sound-design tools.
Producers who want sample editing, synthesis, routing and modulation alongside pattern writing.
Performers who want MIDI-mapped controls and audio-reactive visuals within the same project.
Students exploring sequencing, digital audio, file formats, synthesis and browser-based interfaces.
This guide describes version 5.04. It documents available controls and implemented workflows, not a claim of perfect compatibility with every historical tracker or browser. The screenshots show the real application with an original generated demonstration song; they are not interface mockups.
02 / At a glance
Area
Capacity or architecture
Tracks
Up to 64 tracks; one audio bus and DSP rack per track.
Columns
Up to 4 note/instrument/volume columns and 3 effect columns per track.
Patterns
Up to 256 rows per pattern; sequence positions reference patterns.
Instruments
128 instruments; up to 16 sub-samples per instrument.
Pitch range
C-0 through B-9: 120 distinct notes in MLT. Legacy exports have their own pitch limits.
Envelopes
Volume, panning, filter and pitch; up to 25 points each.
Voice continuation
Cut, Continue, Note-Off and Fade new-note actions; a bounded pool of up to 32 additional NNA voices.
Sample data
Mono or stereo; editable 8-bit/16-bit sample modes, loops, markers and sample processing.
Markers
Up to 256 markers per sub-sample, addressed as hexadecimal 00 through FF.
Racks
Up to 16 DSP blocks per track or master rack. At most 16 automatable parameters per block.
Native devices
41 current DSP and meta-device definitions; detailed controls below.
Synthesis
Two oscillators, mono/stereo generation, unison, waveform library and 128 presets.
Visual Lab
8 effect categories with 16 choices each; four assignable parameter controls per effect.
Interface
Pattern, Instrument, Sample and Mix views; collapsible side panels; 50% to 125% scale.
Project
Self-contained .MLT project storage; legacy module and MIDI interchange, audio and video export.
Note columns share a track's bus and effects. Effect columns have their own memory and a deterministic left-to-right resolution order. More columns do not create independent mixer strips.
03 / Compose and perform
Pattern editing
Enter notes using a QWERTY or AZERTY computer keyboard, or edit hexadecimal values directly. Edit mode separates writing from auditioning.
Select individual fields rather than only whole cells. Note, instrument, volume and command masks allow selective copy, paste and cut.
Move selections, transpose by semitone or octave, interpolate effects and undo pattern edits. Add or remove rows and additional note/effect columns.
Edit track names, insert or remove tracks at a chosen position, and move whole tracks with their racks and associated routing references.
Select a track from its waveform or pattern area to keep the editing cursor and the lower Track FX/Track DSP rack in context.
Use instrument auto-selection, editable instrument names, instrument duplication and reordering, and batch instrument substitutions.
Use the graphical Track FX rack to monitor command values during playback. With Edit mode off it is read-only; with Edit mode on, supported slider/knob changes can be recorded into the pattern. Four-bit commands expose the appropriate nibble range.
The sequence is also a performance surface
Add, delete, copy, paste, duplicate and reorder sequence positions. While stopped, drag a pattern number to change its assignment. Give positions a short label for song structure and Visual Lab text. Green indicators define a loop range; an orange queued position requests the next pattern without immediately abandoning the current one. The timer toggles continuous sequence following.
Transport and timing
Play, Repstart, From, Loop and Stop serve different starting contexts. BPM, Speed and LPB are separate controls: Speed retains tracker tick semantics, while lines-per-beat contributes to musical timing. Start from the editing cursor with Shift + Space. A normal Space stops a running song without switching off Edit mode; when stopped, it toggles Edit mode. Caps Lock enters Note-Off. Shift + Enter toggles fullscreen.
Lock Follow keeps the playback row in view. The pattern grid is virtualized so the interface need not rebuild every cell of a large pattern. Sticky track headers, row numbers, coloured scopes and track meters remain visual reference points.
Pattern commands are part of the instrument
Arpeggios, pitch slides, tone portamento, vibrato, tremolo, retriggers, note delays, sample offsets, envelopes, channel volume, panning and control flow can be written directly into patterns. MLT also has commands for sample-loop manipulation, markers, DSP parameters, MIDI macros and Visual Lab performance. See the complete current helper inventory below.
Composition shortcuts
Pattern block selection requires Edit ON and stopped playback, whether using the mouse or Shift + arrow keys. Switching to Edit OFF clears the range. In Edit OFF, clicks still position the navigation cursor and select the track and its lower rack, without starting a selection.
Shortcut
Action
Shift + Arrow keys
Extend/reduce selection by field or row (Edit ON).
Shift + F3
Cut selected fields.
Shift + F4
Copy selected fields.
Shift + F6
Paste at the cursor or within selected fields.
Shift + F1 / F2
Transpose down/up one semitone.
Shift + F7 / F8
Transpose the selected instrument down/up one semitone.
Shift + Keypad - / +
Transpose down/up one octave.
Shift + Keypad Enter
Audition one row, then advance by Edit Step.
Shift + Space
Start at the editing cursor / stop.
Shift + Main Enter
Toggle fullscreen.
Use either Shift key. Ctrl, Alt, AltGr and Meta must not be held. The Pattern view must be active and song playback stopped.
The Pattern length control in the fixed left dock shows its editable row count above two horizontal arrows: left decreases the length and right increases it. You can also type a value from 1 to 256. Stop playback before resizing a pattern; resizing participates in Undo/Redo.
+Row and -Row are on the static bar below the Pattern editor, immediately before Inst. Auto-select. They insert or remove a row at the editing position; they are separate from changing the pattern's total length.
Without a selection, cut/copy/transpose use the current track in the current pattern, including its visible note/effect columns. They never move DSPs or change other patterns. Paste without a selection starts at the cursor and is clipped at pattern/track boundaries. Only matching field types are pasted. Note / Instrument / Volume / Command masks apply.
Cut, paste and transposition require Edit ON. Notes stay within C-0 through B-9; Note-Off values are not transposed.
Undo and Redo: the buttons beside the editor tabs share one project history with MCP (transaction.undo, transaction.redo and history.read). Stop playback and enable Edit ON before restoring an edit. Up to eight transactions are retained across both directions; a new edit after Undo replaces the Redo branch. A slider gesture is one transaction, rather than one entry per movement.
Edits made during playback remain available after Stop. This includes recorded pattern commands, track and Master DSP adjustments, instrument controls, Mixer faders and panning, and recorded Visual Lab actions. A continuous supported control gesture groups its changes even when playback reaches another pattern. Stop can end a gesture without losing its last committed input. Undo and Redo stay unavailable until playback has stopped.
Playback advancing by itself does not add Undo entries or discard Redo. Only the manual changes and their recorded pattern destinations are retained; voices, playback position and unrelated tick-driven parameter changes are not captured as part of a manual gesture. The same eight-entry and memory limits apply to live edits.
UI edits and permitted live MCP recording use the same chronological history. When a ninth transaction is retained, the oldest is dropped: eight Undo steps then return to the state after that dropped edit, not necessarily the state before playback. Switching between Pattern, Instrument, Sample and Mix does not create another edit. MCP structural edits still require stopped playback; live DSP recording must explicitly request the live destination. After Stop, either the UI buttons or MCP can restore the retained transactions.
Sample settings such as Auto-seek can join the live-edit history. PCM replacement and heavy sample treatments require stopped playback; refusing an unsupported live treatment does not remove earlier edits. After Stop, sample treatments use the same history and retain exact sample data within its memory and storage limits. Undo restores the fields owned by an edit, not every value previously produced by replaying DSP or Visual Lab commands.
After heavy sample editing, you can resume the song and switch editors without committing another edit. Stop again before Undo or Redo. Listening after an Undo keeps the available Redo steps; only a new edit replaces that branch. In-place treatments and operations that resize a sample can share the same timeline, including stereo PCM, loop settings and markers. Eight levels are a maximum: the retained before/after sample data must also fit the shared disk quota, or a further operation can be refused.
T2L, Render synth, Populate and Splitter also join this history when their completed results are committed. Populate restores all affected sample slots and the instrument's keyboard map together. Splitter restores the destination instruments it filled, rather than storing unrelated parts of the song. These operations require stopped playback; you can listen again between Undo/Redo steps without losing the remaining Redo branch. T2L's target duration depends on both the line count and the preview note, so a low-root sample generated by Populate may require fewer target lines to stay within the permitted stretch ratio.
History is temporary and is cleared when a project is replaced. The RAM journal is limited to 32 MiB, so fewer than eight entries may fit. Large fixed-length treatments, sample replacement and resizing use browser storage in 256 KiB pages, with a shared 128 MiB disk limit. This covers trim, cut/paste, mono/stereo and bit-depth conversion, sample loading, synthesis/Populate, TFX, T2L, RP2S, optimization and visual-tool audio exported to a sub-sample. PCM, loop settings, markers and the affected model references are restored together.
A progress dialog blocks editing during preparation and restoration. The affected PCM domain is capped at 64 MiB per state, and the two stored states must fit the 128 MiB disk budget. A whole-bank or whole-project operation can therefore be refused even when its new sample is small. Storage denial, insufficient quota or an oversized operation refuses the edit or restores its original data. Save before large transformations.
Disk history is temporary, not a saved project or a way to resume Undo after reopening MLT. Dropped entries release their pages; abandoned tab data is reclaimed after its lease expires. The PCM owned by the live model and the processing tools' own working buffers are separate from the RAM journal budget. Native project controls include Audio/Perf output settings, instrument colors and AdLib editing; local browser preferences, hardware permissions and playback itself are not restored. Creating or loading another module starts a new history. Save the current module before replacing it.
Song titles, sequence labels and loop marks, sequence pattern changes and range paste/duplication share the same Undo/Redo timeline. So do sample loop settings, Auto-seek, the preview note, pattern-command insertion, MIDI recording preferences, custom LFO curves and DSP impulse-response loading/removal. Selecting a sub-sample, switching editors or changing analyzer display quality does not discard an existing Redo branch.
MIDI library replacement and Drumkit installation use the bounded sample-storage history. The old and new data must fit its limits; a large bank can be refused before installation. Restoring MIDI or Line In settings never grants new device permissions. An unavailable physical input still requires local authorization.
Envelope enable, sustain and loop switches participate in the shared history. Changing the displayed envelope does not discard preceding edits; a checkbox gesture occupies one history entry.
Typing fields, dropdowns, dialogs and Visual Lab fullscreen keep their own keyboard controls. Shift + number keys are not assigned to these shortcuts, so AZERTY numeric entry is preserved. Holding a mutation key does not repeat the modification. Shift + arrow keys can repeat for selection in Edit ON.
Row audition is available in Edit ON or OFF. STOP cancels it. One preparation can run at a time, with an eight-second row limit. It starts from a fresh audio graph: previous held voices, effect memory and DSP tails are not replayed. The nearest preceding instrument in each note column is used when omitted.
Function keys may require Fn on some keyboards. OS-reserved combinations cannot always be overridden by a browser. No other new shortcuts are added.
04 / Instruments and FM
In narrow windows, general instrument controls flow into two columns. Transposition controls and envelope settings wrap below them; scroll the editor vertically to reach the Adlib panel and keyboard. The keyboard has its own horizontal note-range scroll.
MLT 5.04: Instrument editor: general controls, envelope editing, PCM/Adlib configuration and the virtual keyboard.
PCM instruments
Each instrument maps up to 16 sub-samples across the 120-note keyboard. Set main volume and panning, finetune, fadeout, automatic vibrato speed, depth, sweep and waveform. Choose a new-note action to cut, continue, release or fade a previous voice when another note arrives on the same track. Imported legacy instruments use appropriate format-specific defaults rather than assuming polyphony.
Per-note random variationMLT 5.04: Volume, pan, cutoff and resonance variation in knob mode.
The instrument editor provides Volume swing (0-100), Pan swing (0-64, native IT units), Cutoff swing (0-100) and Resonance swing (0-100). All start at zero. Use either sliders or knobs; the four controls share a dedicated row below the general instrument settings.
Each new note draws its own variations. They remain stable while that voice plays, including its later ticks; simultaneous voices do not modify the shared instrument settings. Retriggers retain their current variation. Volume swing scales instrument gain without making a silent instrument audible. Panning is bounded to the stereo range. Filter variations need the corresponding instrument filter controls enabled and do not enable a bypassed filter. Zero base resonance remains zero.
MLT files preserve all four settings. IT export stores volume and pan swing in native instrument fields and cutoff/resonance swing in optional OpenMPT instrument extensions. OpenMPT-aware readers can restore these filter variations; original Impulse Tracker readers cannot reproduce them. IT sample-mode songs switch to instrument mode when needed to retain variations. Standard XM, MOD and S3M cannot store these instrument variations.
Volume and panning envelopes sit alongside MLT's filter and pitch envelopes. Each has up to 25 points, enable state, sustain and loop controls. Filter cutoff and resonance have independent controls; pitch envelopes alter playback pitch. The envelope panels switch type without requiring four permanently expanded editors.
Click the virtual keyboard to audition. Hold Shift while clicking to change sample mapping. Coloured key outlines identify played instruments, and envelope playheads show active voices. Instrument-list panning controls are linked to the main instrument panning control.
Adlib / OPL instruments
MLT includes FM instrument editing for Adlib material, including S3M instruments. The modulator and carrier expose operator level, multiplier, key scaling, envelopes, waveform and modulation flags, with connection and feedback controls. An Adlib-enabled instrument replaces its PCM sound source, rather than layering the two accidentally.
The GENMIDI bank supplies 128 General MIDI melodic presets and a percussion selection. FM is an alternative timbre source, not a sampled imitation of every acoustic instrument. Hardware-perfect OPL reproduction across every module is not claimed here.
Native S3M FM instruments use the Opal OPL3 engine, including their eight waveform types, tracker tuning and supported channel panning. Each instrument still feeds its track's DSP rack. The existing OPL2 and GENMIDI instrument controls remain available.
General MIDI libraries
Load a local GM.DLS or SF2 bank, use the downloadable GeneralUser-GS bank, or choose the FM bank. DLS articulations and sample-region information are converted into MLT instrument settings where supported. Drumkit generation maps percussion into sample slots. Bank conversion is constrained by MLT's instrument and sample-map capacities.
MLT 5.04: MIDI library controls: local-bank and General MIDI instrument workflows.
A browser cannot silently read a system GM.DLS file. Local bank selection needs user action. The GeneralUser-GS download uses a pinned resource with verification and caching; its licence and attribution remain relevant when redistributing bank content.
Edit mono or stereo samples with frame-level selection, zoom and editable loop boundaries. Forward and ping-pong loops have draggable start/end handles with bounds checking. Convert between mono and stereo, change sample bit depth, preview at a reference note, and undo sample changes.
Task
Tools
Editing
Cut, copy, paste, mix, trim, selection and waveform drawing.
Level and shape
Normalize, apply volume, linear/logarithmic fades, reverse, smooth and invert polarity.
Loops
Loop selection, forward/ping-pong playback, boundary dragging and loop crossfade.
Timing
T2L stretches to a chosen number of lines using song timing and reference pitch.
Resampling effects
TFX applies a chosen native track rack to sample material after checking unsupported sources and render limits; RP2S renders the current pattern into a stereo 16-bit sub-sample.
Silence
Insert silence or create a new silent sub-sample with a duration based on pattern rows.
Markers
Add/delete, drag, number, preview and stop markers; detect transients with adjustable sensitivity.
Pattern links
C2P writes the selected marker command; the offset ruler can write sample-offset commands in Edit mode.
Export
Save the selected sub-sample as WAV.
Two offset systems
The 9xx ruler follows tracker sample-offset addressing; the high-offset command can extend it when another effect column is available. The Sxx ruler spans 00 to FF. With markers, Sxx addresses a marker. Without markers, the fallback offset is floor(sampleFrames × xx / 256). Thus S40 starts one quarter of the way through an unmarked sample, not at frame 64.
Long samples and Auto-seek
Per-sample Auto-seek helps From/cursor-start reconstruct a long PCM sample that began earlier in the song. Reconstruction follows tracker state rather than merely subtracting elapsed seconds. It is bounded and cancellable. It does not promise a fully warmed history for every delay, reverb, external input or OPL voice.
Transient detection and transformation
Detect transients identifies attacks and creates sample markers, with adjustable sensitivity and a maximum of 256 markers. It does not identify musical pitches or transcribe a sample into notes.
SPLITTER provides spectral-focus separation and duration/memory controls. It is not a trained neural stem-separation service. Keep source duration and memory budget suitable for the machine; splitting cannot perfectly isolate every voice in a finished mix.
T2L stretches the sample to the number of lines in its adjacent field. The target uses song BPM, Speed, LPB timing and the selected preview note, including the sample's tuning. It is the stored frame count needed to play that duration at the preview note, not simply the duration at the sample's original rate.
MLT uses the real Signalsmith Stretch WebAssembly engine, with dense offline analysis/synthesis overlap rather than repeating grains or processing stereo channels independently. Left and right share one engine and one timeline. Durations remain exact after latency compensation. The 1x case copies the original PCM unchanged.
While processing, the T2L button becomes Cancel and shows progress. Click it to abort without changing the sample or the previous Undo. Completed stretching can be undone, including both channels and the original loop boundaries. Changing the selected sample during processing prevents the old result from overwriting the new selection.
Calculations run in a disposable Worker, leaving the interface available. A conservative 128 MiB working-memory budget and a 0.05x-20x duration-ratio limit are checked before PCM copies or output allocation. Very large requests are refused explicitly. The high-quality overlap costs more processing time, particularly in Firefox, but does not run on the interface thread.
Extreme stretching can soften attacks or change the texture. Preview the result on your material; Undo restores the previous sample.
Render track effects into a sample (TFX)
TFX applies the selected track's enabled audio DSPs, in their rack order, to the selected sample range or the whole sub-sample. It now builds the same native DSP graph used by the track rack, including its current parameters and imported impulse responses. It does not use a separate collection of simplified effects.
MLT 5.04: Review the chain before committing an edit. A refused or cancelled operation preserves both the sample and the previous Undo.
Stop song playback, choose the source track beside TFX and select a sample range if needed. Click TFX to inspect the chain, then Render.
All 32 autonomous native audio devices are supported. Disabled blocks are bypassed. Rack collapse state has no effect on processing.
Tail seconds is zero by default, preserving the current length and trimming decay. Up to 10 seconds can be appended when the selection reaches the sample end. An interior selection cannot shift the untouched material after it.
Stereo input stays stereo. A mono sample becomes stereo when the rack produces different left and right signals. Unselected PCM remains unchanged; conversion to the sample's existing bit depth happens only once, at the end.
The render excludes the master rack, channel fader, AGC, instrument pitch/envelopes and pattern automation. It uses a fresh DSP state at the stored sample rate. Inertia does not recreate an earlier song state.
The memory estimate is checked before audio allocation. The default is 96 MiB, adjustable from 16 to 256 MiB. A shorter selection or tail reduces the render cost, although the final sample replacement and Undo still need memory.
Overload is reported before writing. Lower the rack output, or explicitly enable Allow PCM clipping. TFX does not silently normalize or clip between DSPs. Undo restores the prior PCM, channel layout and loop metadata.
When TFX refuses a chain
Router, Vocoder, Voice Carver and Line In need other tracks or a live source. The five audio meta devices need song routing or note/timeline context. An enabled instance blocks the operation with an explanation, rather than silently disappearing from the result. Disable those devices deliberately, or render the complete song. Meta Visualizer changes no audio and is explicitly excluded.
Auto-tune, Reverse Delay and LoFi Mat now run their original processing loops in a shared AudioWorklet for live and offline use. This also fixes missing wet audio in Firefox offline rendering. Fixed lookahead in the Maximizer and multiband devices, plus Auto-tune's block adapter delay, is compensated by TFX. Other intrinsic or parallel-path delays remain part of the effect.
Pitch detector
Open Pitch detector in the Sample editor to estimate notes from its selected frame range, or the whole sample. Choose one to four voices, confidence threshold and minimum note duration, then Detect notes. Progress and Cancel remain available while a worker processes up to 524288 frames. Results show note names, timing and confidence; click a note to select its waveform range. Analysis never writes pattern notes or modifies PCM.
This is a harmonic spectral estimator shared with the command API, not the Basic Pitch neural model. Stereo channels are analyzed independently to avoid antiphase cancellation. Noisy sounds, transients, dense chords and very low fundamentals can be ambiguous; inspect results before using the separate command-based insertion workflow.
06 / Build sounds in the sample editor
Synth controls use as many columns as the available width permits, in both knob and slider modes. Scroll the Sample editor vertically to reach the complete synth; all parameters remain available.
The Mono Synth becomes a Stereo Synth for stereo material. Two oscillators combine familiar primitive waveforms with imported AKWF waveforms. Oscillator previews make the source shape visible. The AKWF browser supports browsing, auditioning and assigning library waves to either oscillator.
Oscillator balance, detuning and noise; filter mode, cutoff, resonance and envelope influence.
Amplitude ADSR, LFO rate and pitch/filter modulation, and drive.
Up to seven unison voices, stereo spread, drift and oscillator panning for width beyond duplicated mono.
Slider or knob editing, keyboard-armed audition and selectable generation quality.
Frame-length control for short looping cycles or substantially longer sounds; existing selection/sample duration can supply the starting length.
Preview puts generated audio on the sample clipboard. Render to Slot mixes into a pre-existing selection; without a selection it renders the sample content. Undo provides a way back.
Populate generates and maps 16 sub-samples across the instrument keyboard. This replaces the instrument's sample population, so preserve a copy before using it.
Loop preparation and crossfading reduce boundary clicks. Very short cycles and extreme processing can still change the perceived spectrum or expose discontinuities; audition the loop at the intended pitch.
07 / Context-aware arrangement
MLT 5.04: The arrangement assistant in the right panel, alongside the tracker and its current musical context.
The assistant generates Lead, Bass, Chords, Drums or Ensemble proposals. It can use notes in the current pattern and preceding sequence positions, follow the harmony, avoid clashes or ignore context. Key, scale, style, density, register, complexity, seed and variation shape the result.
Its approach combines rules and templates, shared harmonic planning, bounded voice-leading search, motifs, bass/rhythm relationships, fills and optional Euclidean rhythm ideas. It is not an unrestricted neural composition engine. A repeatable seed makes a proposal reproducible; the seed randomizer explores alternatives.
Generate, inspect, audition, vary, apply and undo. Preserve or replace material within the chosen pattern/selection and field masks. Applying requires stopped playback and Edit mode. The available palette includes 21 scale collections and 12 style families; these are starting points, not a promise to model every musical genre.
08 / Mixing, routing and automation
Mixer strips keep a fixed usable width. Scroll horizontally for later tracks and scroll each DSP matrix vertically for longer chains. The lower half of each strip remains reserved for meters, channel volume and panning.
MLT 5.04: Mix view: channel strips, DSP matrices, stereo meters, channel-volume faders and pan controls.
The mixer presents named tracks with Mute, Solo and Surround, ordered DSP blocks, stereo RMS meters, channel-volume faders and panning. Channel volume ranges from 0 to 64, or hexadecimal 00–40, matching the channel-volume role rather than a 0–128 global-volume scale.
Select anywhere in a strip to change the active track. Edit its name, swap adjacent devices with the diamonds, or drag a device to a precise rack position. Cross-track dragging and arrow moves transfer a device when the destination is valid. Destination feedback prevents duplicate or impossible installations; associated command and meta-routing references are updated. A successful structural move may stop transport for a safe reroute.
MLT 5.04: Track DSP rack: native effects, controls, presets and the selected analysis surface.Song Levels: initial volume and source preamps
Open SONG & MEDIA on the left and scroll to SONG LEVELS, immediately above OPTIMIZE. These settings belong to the song, not the final Master output.
Setting
Use
Initial Global Volume
0–128. Sets the starting volume for song playback and stopped instrument audition. Editing this initial value does not override a global-volume command already running in the song. Pattern commands can subsequently change the runtime global volume without rewriting the saved initial value.
Sample Volume
0–2000. Controls the PCM preamp before track DSPs. Imported songs retain their format's default mix policy; new native songs use 128 for unity PCM gain. This is not the volume of one instrument, nor the destructive Apply Volume operation in Sample Editor.
Synth Volume
0–2000; 48 is unity. Controls FM/AdLib output after chip synthesis and before track DSPs, without changing the operators' modulation or timbre. It does not attenuate PCM. Sounds rendered to slots by Mono/Stereo Synth are samples and therefore use Sample Volume.
Zero mutes the corresponding source; high preamp values can clip. Master Volume and AGC are separate output controls. PCM and Synth preamp edits are consumed by the normal playback/preview scheduling; no additional refresh timer is used.
Independent song and source levels in the left panel.
File compatibility: MLT preserves all three values. IT stores initial global volume and sample preamp in its header; OpenMPT-compatible song extensions retain Synth Volume, preamps beyond the native header range and finer initial-volume values. XM uses OpenMPT-compatible extensions for these settings; classic FT2 players may ignore them. S3M retains its supported initial-volume and sample-preamp fields, with legacy quantization and hardware-profile limits, but has no separate Synth Volume field. MOD has no native storage for these song-level settings. Export warns before incompatible MOD/S3M delivery; keep an MLT or compatible IT copy when these levels matter. No automatic PCM baking is performed.
LFO, Signal Follower, Velocity Tracker, Key Tracker and Cerberus route modulation through a small source-to-target model. Routes point to eligible devices to the right in the same track or to later tracks, avoiding direct cycles. Invalidated targets can become Undefined. Cerberus has three destinations. Meta Visualizer controls Visual Lab selection, enable state and parameter bindings.
Pattern automation
Direct four-hex-digit track DSP commands identify a block, a parameter and an 8-bit value. Master-rack focus/value commands use I and J. Automation indexes span 0–F, with a maximum of 16 automatable parameters; a device can have additional non-automatable controls. Inertia changes transition speed: lower means smoother/slower movement, higher means faster changes. Router and Meta Visualizer do not add Inertia.
Audio routing and sidechains
Router sends a track's audio to a later track, optionally muting its direct post-effects output. Vocoder 24 uses the current track as carrier and a preceding router as modulator. Voice Carver performs 24-band sidechain spectral reduction. Line In mixes a permitted capture source into a track; a source cannot be assigned redundantly across Line In devices.
Amplify uses the active tracker mixing convention. Native FT2 playback retains its x1-x32 scale (x10 by default), without an additional default sample preamp. Imported IT, S3M, MOD and OpenMPT-compatible XM profiles retain their stored mix levels; x10 is unity output trim after those levels. Explicit Sample Volume and Mixer volume settings remain effective. OpenMPT and FT2 can have different reference loudness even with unchanged notes. This is not peak normalization; higher amplification can still overload a loud mix. Use the AGC or a limiter when needed.
09 / Every native device
This inventory is extracted from the running 4.80 device definitions, not from historical requests. Expand a device for its controls, ranges, automation indexes and current preset names. Some targets and parameter labels are context-dependent; a meta-device target selector is populated by eligible installed devices.
Stereo-linked, phase-compensated three-band compression without automatic makeup; reconstructed-peak protection follows Output and Mix.
Control
Range / choices
Default
Automation index
Low X
60 to 1200 Hz
180
0
High X
1200 to 12000 Hz
4200
1
Low Threshold
-60 to 0 dB
-18
2
Mid Threshold
-60 to 0 dB
-16
3
High Threshold
-60 to 0 dB
-14
4
Low Ratio
1 to 20 :1
2
5
Mid Ratio
1 to 20 :1
1.8
6
High Ratio
1 to 20 :1
1.6
7
Attack
0.1 to 100 ms
18
8
Release
20 to 2000 ms
260
9
Soft Knee
0 to 18 dB
9
A
Final Ceiling
-6 to 0 dBTP
-1
B
Release Shape
Uniform / Frequency / Program
Frequency
C
Mix
0 to 100 %
100
D
Output
-18 to 12 dB
0
E
Inertia
0 to 100 %
72
F
Presets
Flat Three-Band
Transparent Master
Low-End Control
Midrange Focus
High-End Tamer
Vocal Balance
Drum Bus Punch
Bass Tightener
Broadcast Control
Gentle Glue
AI Band Balance
AI Harshness Control
AI Mud Cleanup
Dense Mix Recovery
Dense Band Control
Three-Band Safety
Clean multiband dynamics
The 3-Band Compressor and Multiband Limiter now use explicit stereo-linked gain reduction, phase-compensated Linkwitz-Riley crossovers and a final lookahead peak guard. There is no browser-compressor makeup gain and no output waveshaper. Quiet signals retain their level; compression deliberately changes the balance only when a band exceeds its threshold.
MLT 5.04: Scroll the rack horizontally to reach the final controls. The new compressor ceiling replaces automatic makeup.
Disable AGC while assessing either processor. To protect the complete song, place the device last in the Master rack. Effects or gain stages after it can create new overloads.
Low X and High X set the split frequencies. Each band has its own threshold; the compressor also has independent ratios. The limiter's band thresholds are not independent output guarantees: Final Ceiling protects their summed result.
Mix blends the dry and reduced versions of the same phase-aligned bands. Mix 0% retains the crossover phase and final protection. Disable the device for actual bypass.
Output is explicit gain before final protection. The compressor no longer adds automatic makeup. Increase Output deliberately if compression needs level compensation.
Final Ceiling is protected after summing, mixing and output gain, using the Maximizer's 4x reconstructed-peak detector and conservative 0.8 dB safety margin. Latency is about 8.67 ms at 48 kHz. No new rack-wide delay compensation is provided.
Inertia remains automatable. The compressor has 16 automation targets; the limiter has 15. Sixteen named presets per device are retained and revised.
The same processing is used for live playback, WAV, RP2S and sample-editor TFX with Undo. TFX removes the fixed processing delay and preserves the selected length. Aggressive gain reduction can still pump or alter dynamics; this is not a denoiser, loudness normalizer or a true-peak meter. A crossover has unity magnitude when recombined, but is not linear phase.
3D Output 14 controls · 14 automation targets
Places this track in the MLT spatial field.
Control
Range / choices
Default
Automation index
Mode
MLT HRTF / Orbit HRTF / Equal-power
MLT HRTF
0
Azimuth L/R
-180 to 180 deg
0
1
Elevation
-90 to 90 deg
0
2
Distance
0.2 to 12 m
1.2
3
Width
0 to 220 %
100
4
Focus
0 to 100 %
55
5
Reflect
0 to 100 %
18
6
Room
1 to 18 m
4.5
7
Material
Anechoic / Curtain / Studio / Wood / Concrete
Studio
8
Height Cue
0 to 180 %
120
9
Rear Cue
0 to 200 %
150
A
Air
-10 to 10 dB
0
B
Mix
0 to 100 %
100
C
Inertia
0 to 100 %
72
D
Presets
Front Center
Close Front
Far Front
Close Rear
Far Rear
Above Head
Below Listener
Wide Front
Wide Rear
Small Room
Reflective Room
Absorptive Room
Headphone Natural
Speaker Natural
AI Spatial Relief
3D Translation Check
3D Output: orbit the listening-space view
Click the cube to rotate its view by one step, or drag horizontally and vertically to orbit. Double-click restores the initial view. Enter or Space also rotates the focused cube. Camera rotation is visual only: source azimuth, elevation, distance and the audio output remain unchanged.
Native DSPs are built into MLT. This is not a VST/AU plug-in host. HRTF positioning is a stereo/headphone technique, IT Surround uses polarity relationships, and neither implies a physical multichannel speaker output.
Alligater 13 controls · 13 automation targets
Tempo-locked stereo rhythmic gating with click-safe curved edges, swing and sixteen patterns. Not a threshold noise gate.
Control
Range / choices
Default
Automation index
Clock
Free Hz / 1/4 / 1/8 / 1/16 / 1/32 / 1/8 triplet
1/16
0
Free rate
0.05 to 40 Hz
4
1
Pattern
Four on floor / Eighths / Sixteenths / Offbeat / Trance / Tresillo / Syncopated / Breaks / Sparse / Gallop / Alternating / Half bar / Late surge / Staccato / Clusters / Closed
Sixteenths
2
Steps
4 / 8 / 16
16
3
Gate
5 to 100 %
55
4
Attack
0.1 to 50 ms
3
5
Release
0.1 to 150 ms
12
6
Depth
0 to 100 %
100
7
Swing
0 to 75 %
0
8
Stereo shift
0 to 1 steps
0
9
Mix
0 to 100 %
100
A
Output
-18 to 12 dB
0
B
Inertia
0 to 100 %
72
C
Presets
Clean sixteenths
Soft eighths
Offbeat chop
Trance pulse
Swung sixteenths
Tresillo motion
Syncopated pad
Broken rhythm
Wide ping gate
Stereo offbeat
Sparse breaths
Galloping lead
Alternating bars
Half-bar swell
Late surge
Cluster texture
Alligater: rhythmic amplitude gating
Alligater is a rhythmic gate, distinct from the threshold-based Gate. Sixteen named patterns, 4/8/16-step lengths, musical or free-rate timing, duty, swing and stereo phase offset shape a gain envelope. Attack and Release soften its edges. Depth and linear Dry/Wet preserve explicit control over attenuation; Output is the only deliberate level increase. Inertia smooths continuous control changes. The device has thirteen automation targets including Inertia and sixteen presets.
Live racks follow scheduled tempo ticks. TFX uses the same native processor at the current or explicitly supplied fixed BPM, without AGC or normalization. It does not replay song tempo automation. Depth zero or fully dry Mix preserves the input at 0 dB Output.
Autotune 14 controls · 13 automation targets
YIN pitch tracking with confidence gating, scale hysteresis, transient protection and three-grain pitch correction.
Control
Range / choices
Default
Automation index
Key
C / C# / D / D# / E / F / F# / G / G# / A / A# / B
C
0
Scale
Chromatic / Major / Minor / Harmonic Minor / Melodic Minor / Dorian / Mixolydian / Pentatonic
Valve-inspired preamp and push-pull power stage with dynamic bias/supply sag, then short cabinet convolution. Built-in cabinets are synthetic models; load a measured mono/stereo WAV IR for a captured speaker/microphone response. Cabinet Only bypasses the amplifier. Quality and cabinet selection are manual settings.
Control
Range / choices
Default
Automation index
Amplifier
Clean combo / British crunch / Cascaded lead / Cabinet only
British crunch
0
Preamp
0 to 42 dB
16
1
Bass
-12 to 12 dB
0
2
Middle
-12 to 12 dB
0
3
Treble
-12 to 12 dB
0
4
Presence
-9 to 9 dB
0
5
Power
0 to 100 %
35
6
Supply sag
0 to 100 %
35
7
Cabinet
Model: open 1x12 / Model: closed 4x12 / Model: bright 2x12 / Model: bass 4x10 / Loaded WAV IR / No cabinet
Model: closed 4x12
Manual
Off-axis
0 to 100 %
25
8
Low cut
20 to 250 Hz
35
9
High cut
2000 to 18000 Hz
14000
A
Oversample
2x economical / 4x high
4x high
Manual
Mix
0 to 100 %
100
B
Output
-30 to 6 dB
-2
C
Inertia
0 to 100 %
72
D
Presets
Edge of breakup
Open-back clean
Glass chime
Warm jazz combo
Small combo pushed
British rhythm
Singing lead
Tight high gain
Vintage power bloom
Dark doom stack
Bass cabinet clean
Bass parallel grit
Synth stack
Drum room crunch
Speaker only
External cabinet feed
Cabinet Simulator: preamp and impulse responses
An amplifier is more than a tone filter, and a cabinet impulse response is not an amplifier. This native track/Master device combines nonlinear, valve-inspired preamp stages, dynamic bias, a push-pull power-stage approximation and supply sag with short cabinet convolution.
MLT 5.04: Current 5.04 track rack, knob view at 100% scale. Scroll the rack horizontally for cabinet, microphone, mix, output and Inertia controls.
Start here
Enable Edit mode, open a Track DSP or Master rack, and choose Cabinet Simulator in its helper.
Enable the inserted block using its bypass/power control. Start with Open-back clean, British rhythm or Singing lead.
Use Preamp for preamp drive, Power for power-stage drive, Supply sag for dynamic compression, and Inertia for parameter transition speed. Compare at similar output levels.
Choose one of the four built-in cabinet models. For a captured speaker/microphone response, stop playback and use Load IR with a mono/stereo WAV of up to 500 ms. Clear IR restores a built-in model.
Choose Cabinet only when an upstream source already supplies the amplifier sound. Save as .MLT to retain the DSP and embedded IR.
The 16 presets cover clean combos, chime, jazz, crunch, lead, bass, synths, drums and cabinet-only processing. Oversample offers 2x economical and 4x high quality. There are 16 controls, including 14 automation targets; cabinet choice and oversampling are manual. Presets retain a loaded IR asset, but may select a built-in model instead.
Native browser audio nodes process the signal, with no per-sample JavaScript. Built-in IR kernels are cached, changes crossfade, and removal disposes the audio graph. Imported IRs are bounded and stored inside native projects. Hardware-rate changes use windowed-sinc IR resampling.
The built-in cabinets are original synthetic models, not measured commercial captures. The amplifier is a reduced-order valve-inspired model, not an exact clone of a particular physical amplifier. An external IR supplies a linear cabinet response; it does not capture nonlinear valve dynamics. Use the output meters: high-gain input or strong IRs can clip.
Cerberus 17 controls · 7 automation targets
Routes one meta source to three installed DSP parameters on this track.
Control
Range / choices
Default
Automation index
Source Meta
Meta LFO / Signal Follower / Velocity Tracker / Key Tracker
Free rate / Gate / Attack / Release / Depth / Swing / Stereo shift / Mix / Output / Inertia / Gain / Panning / Offset / HP Cutoff / Low Gain / Mid Gain / Mid Freq / Mid Q / Hi Gain / 31 / 63 / 125 / 250 / 500 / 1k / 2k / 4k / 8k / 16k / Q / 60 / 12k / Azimuth L/R / Elevation / Distance / Width / Focus / Reflect / Room / Height Cue / Rear Cue / Air / Frequency / Drive / Harmonics / Color / Peak Guard / Air Limit / Amount / Carrier Note / Transpose / AM Bias / Stereo Phase / Low Cut / Anti-Alias / Tone / Bit Depth / TPDF Dither / Noise Shape / Clock Jitter / Reconstruction / Stereo Link / L Delay / R Delay / L Feedback / R Feedback / Send / Thresh / Hold / Floor / Feedback / Delay / Phase / Amplitude / Filter A / Filter B / Filter C / Ceiling / Bass Width / Bass Split / Decorrelation / Mono Safety / Balance / Damping / Stereo Spread / Resonance / Ripple / Low X / High X / Low Threshold / Mid Threshold / High Threshold / Low Ratio / Mid Ratio / High Ratio / Soft Knee
Manual
Amount C
-100 to 100 %
0
3
Min
0 to 100 %
0
4
Max
0 to 100 %
100
5
Inertia
0 to 100 %
72
6
Presets
Default
Preset 02
Preset 03
Preset 04
Preset 05
Preset 06
Preset 07
Preset 08
Preset 09
Preset 10
Preset 11
Preset 12
Preset 13
Preset 14
Preset 15
Preset 16
Chorus 7 controls · 7 automation targets
Short modulated delay for widening.
Control
Range / choices
Default
Automation index
Rate
0.01 to 12 Hz
0.65
0
Depth
0.1 to 30 ms
8
1
Feedback
-95 to 95 %
8
2
Delay
1 to 60 ms
18
3
Dry/Wet
0 to 100 %
35
4
Phase
0 to 360 deg
90
5
Inertia
0 to 100 %
72
6
Presets
Subtle Chorus
Classic Chorus
Wide Chorus
Slow Ensemble
Fast Ensemble
Vocal Double
Guitar Chorus
Synth Chorus
Bass-Safe Chorus
Dark Chorus
Bright Chorus
Deep Modulation
AI Width Restore
AI Static Relief
Vintage Ensemble
Chorus Phase Check
Comb Filter 10 controls · 9 automation targets
Stereo tuned feedback comb with fractional delay, damping, drive, polarity and width control.
Control
Range / choices
Default
Automation index
Note
0 to 119
48
0
Transpose
-48 to 48 st
0
1
Feedback
-98 to 98 %
32
2
Damping
200 to 20000 Hz
12000
3
Stereo Spread
-50 to 50 cent
0
4
Drive
0 to 18 dB
0
5
Dry/Wet
0 to 100 %
45
6
Output
-18 to 6 dB
0
7
Quality
Eco / 2x / 4x
2x
Manual
Inertia
0 to 100 %
72
8
Presets
Neutral Resonator
C2 Wooden Resonance
C3 String Resonance
C4 Metallic Resonance
Negative Feedback Hollow
Wide Stereo Resonator
Dark Damped Comb
Bright Tight Comb
Percussion Body
Plucked String Body
Robot Formant
Short Metallic Room
AI Body Restore
AI Static Breakup
Extreme Resonance
Comb Safety Check
Compressor 10 controls · 10 automation targets
Stereo-linked compression without automatic makeup or coloration; SC HPF filters only the detector. Attack permits transients: use a limiter for a strict ceiling.
Control
Range / choices
Default
Automation index
Threshold
-80 to 0 dB
-18
0
Ratio
1 to 20 :1
4
1
Attack
0.1 to 100 ms
10
2
Release
1 to 500 ms
80
3
Knee
0 to 8 dB
2
4
Slew
1 to 150 ms
20
5
SC HPF
20 to 500 Hz
60
6
Makeup
0 to 30 dB
0
7
Mix
0 to 100 %
100
8
Inertia
0 to 100 %
72
9
Presets
Default Control
Transparent Leveler
Vocal Natural
Bass Firm
Drum Punch
Bus Glue
Parallel Smash
Peak Tamer
Soft Knee Master
Fast Percussion
Slow Program
Podcast Leveler
AI Dynamics Repair
AI Pump Reduction
Dense Mix Control
Strong Compression
Compressor: transparent stereo-linked dynamics
The simple Compressor reduces level above Threshold using Ratio, a soft Knee and Attack/Release smoothing. Slew controls detector-power smoothing in milliseconds. SC HPF filters only the detector: bass remains in the audible signal. Both channels use the same gain reduction to preserve the stereo balance.
There is no automatic makeup, saturation or clipping stage. Ratio 1:1 with Makeup 0 dB is unity; signals below the knee are unchanged. Makeup is an explicit gain increase on the compressed branch. Mix blends aligned dry and compressed signals with linear weights. A low Mix deliberately retains more uncompressed signal.
Attack and detector smoothing allow initial transients through. A compressor is not a ceiling limiter: for guaranteed peak control, place a Maximizer or Limiter after it. Disable AGC when comparing level, compression or transparency; otherwise AGC can hide the result.
Live racks, song rendering and TFX use the same canonical processor and parameter interpretation. TFX waits for initialization and releases the processor after rendering; it does not normalize the result or add a second compressor. Static TFX starts with a fresh detector, unlike a continuously running song. Inertia smooths subsequent parameter changes without rebuilding the graph.
MLT 5.04 simple Compressor with ten controls including InertiaDC Offset 6 controls · 6 automation targets
Adds or removes DC offset on the signal.
Control
Range / choices
Default
Automation index
Mode
Remove DC / Add DC
Remove DC
0
Offset
-25 to 25 %
0
1
HP Cutoff
2 to 30 Hz
8
2
Mix
0 to 100 %
100
3
Output
-12 to 6 dB
0
4
Inertia
0 to 100 %
72
5
Presets
Remove DC 5 Hz
Remove DC 8 Hz
Remove DC 12 Hz
Remove DC 20 Hz
Gentle DC Block
Full DC Block
Parallel DC Block
Low-End Safe Block
Add Positive 1%
Add Negative 1%
Add Positive 5%
Add Negative 5%
Calibration Positive
Calibration Negative
Analyzer Test
DC Safety Check
Delay 8 controls · 8 automation targets
Stereo delay with feedback and send control.
Control
Range / choices
Default
Automation index
L Delay
1 to 2000 ms
180
0
R Delay
1 to 2000 ms
240
1
L Feedback
0 to 96 %
22
2
R Feedback
0 to 96 %
22
3
Send
0 to 100 %
24
4
Line Sync
Off / On
false
5
Mute Source
Off / On
false
6
Inertia
0 to 100 %
72
7
Presets
Short Stereo
Slapback
Quarter Echo
Dotted Echo
Wide Stereo Echo
Left Lead Echo
Right Lead Echo
Soft Feedback
Long Feedback
Dry Tap
Wet Tap
Dub Echo
AI Space Builder
Transition Echo
Rhythmic Stereo
Delay Safety Check
Distortion 15 controls · 13 automation targets
Two-stage pedal distortion: silicon diodes, feedback overdrive, cascaded fuzz and LED crunch, plus digital Shape/Razor/Fold/Shift. Dynamic bias, supply sag, tightness, tone and body. Model and oversampling are manual; use Cabinet Simulator after this device for a speaker response.
Distortion now provides four pedal-inspired families: Silicon drive, Feedback overdrive, Cascaded fuzz and LED crunch. Shape, Razor, Fold and Shift remain available for digital sound design. Two nonlinear stages, frequency-dependent drive, asymmetric junctions, dynamic bias and supply sag shape the response to the input rather than merely changing its level.
MLT 5.04: Actual knob view at 100% scale. Scroll horizontally for the remaining controls. Both sliders and knobs were checked from 50 to 125%.
Enable Edit mode, insert Distortion from the Track DSP or Master helper, and enable the block.
Start with Silicon rhythm, Mid-forward drive, Sustaining fuzz or Open LED crunch. Lower the source level to hear how the distortion cleans up; raise Drive for stronger clipping.
Tightness reduces bass entering the clipping stages. Pre Tone changes which frequencies drive them; Body and Post Color shape the output. Dynamic Bias adjusts signal-dependent asymmetry. Supply sag softens level and headroom under load.
Gain Comp. reduces the calibrated level increase without reducing drive itself. Inertia smooths automated movements. Circuit and Quality are manual choices; the other 13 controls are automation targets.
For a speaker/microphone sound, place Cabinet Simulator after Distortion and choose Cabinet only. A pedal is not a speaker simulator; a full amplifier after it creates additional saturation.
There are 16 new presets and native 2x/4x oversampling. Circuit changes crossfade; rapid changes coalesce. Sample-editor TFX uses this same stereo audio graph with Undo, not the earlier simplified waveshaper. TFX reserves an additional PCM working budget of 96 MiB and rejects an oversized selection before starting this render.
Level-matched comparison WAV: previous default at 0 s; Silicon rhythm at 4.5 s; Mid-forward drive at 9 s; Sustaining fuzz at 13.5 s; Open LED crunch at 18 s; Pedal into Cabinet at 22.5 s. The source is an original synthetic pluck, not a guitar recording.
These are circuit-inspired models, not exact replicas of branded pedals. Adjust input drive and output level to suit the source.
DSP Filter 10 controls · 7 automation targets
Unified analog and digital filter with native cascaded sections, smooth automation and oversampled nonlinear analog stages.
Morphs through sixteen vowel resonances with controllable articulation and transition speed.
Control
Range / choices
Default
Automation index
Vowel
0 i / 1 y / 2 u / 3 e / 4 close oe / 5 o / 6 schwa / 7 epsilon / 8 nasal epsilon / 9 oe / A nasal oe / B open o / C nasal open o / D a / E back a / F nasal back a
0 i
0
Clarity
0 to 100 %
72
1
Mix
0 to 100 %
100
2
Output
-18 to 18 dB
-1
3
Inertia
0 to 100 %
72
4
Presets
Easy Formants Default
Easy Formants Transparent
Easy Formants Gentle
Easy Formants Warm
Easy Formants Bright
Easy Formants Deep
Easy Formants Tight
Easy Formants Wide
Easy Formants Narrow
Easy Formants Rhythmic
Easy Formants AI Cleanup
Easy Formants AI Restore
Easy Formants Mix Polish
Easy Formants Creative
Easy Formants Strong
Easy Formants Safety Check
EQ 5 9 controls · 9 automation targets
Five-band graphic equalizer.
Control
Range / choices
Default
Automation index
60
-18 to 18 dB
0
0
250
-18 to 18 dB
0
1
1k
-18 to 18 dB
0
2
4k
-18 to 18 dB
0
3
12k
-18 to 18 dB
0
4
Q
0.35 to 4
1
5
Field
LR / L / R / M / S
LR
6
Output
-12 to 12 dB
0
7
Inertia
0 to 100 %
72
8
Presets
Flat Five-Band
Warm Tilt
Bright Tilt
Bass Tight
Low-Mid Clear
Vocal Forward
Presence Soft
Air Lift
Drum Punch EQ
Synth Focus EQ
Radio Curve
Club Curve
AI Broad Balance
AI Dullness Repair
AI Harshness Repair
Five-Band Check
Exciter 13 controls · 12 automation targets
Adds only the generated high-frequency harmonic residue while preserving the original signal and its transients.
Control
Range / choices
Default
Automation index
Frequency
800 to 16000 Hz
4200
0
Slope
12 dB/oct / 24 dB/oct
24 dB/oct
1
Character
Smooth / Balanced / Odd / Even
Smooth
2
Drive
0 to 18 dB
6
3
Harmonics
0 to 100 %
32
4
Color
-100 to 100 %
10
5
Peak Guard
0 to 100 %
45
6
Air Limit
6000 to 22000 Hz
18000
7
Quality
Eco / Normal / High
Normal
Manual
Amount
0 to 100 %
18
8
Mix
0 to 100 %
100
9
Output
-12 to 6 dB
0
A
Inertia
0 to 100 %
72
B
Presets
Exciter Default
Exciter Transparent
Exciter Gentle
Exciter Warm
Exciter Bright
Exciter Deep
Exciter Tight
Exciter Wide
Exciter Narrow
Exciter Rhythmic
Exciter AI Cleanup
Exciter AI Restore
Exciter Mix Polish
Exciter Creative
Exciter Strong
Exciter Safety Check
Flanger 12 controls · 10 automation targets
Stereo through-zero-style flanger with phase-offset modulation and a selectable post-modulation color processor.
Control
Range / choices
Default
Automation index
Amount
-100 to 100 %
42
0
Rate
0.01 to 12 Hz
0.28
1
Amplitude
0.1 to 12 ms
3
2
Feedback
-95 to 95 %
36
3
Delay
0.1 to 20 ms
4
4
Phase
0 to 360 deg
180
5
Filter
Off / AM Sine / AM Triangle / AM Saw / AM Pulse / Distortion Shape / Distortion Fold / Decimator / Comb / Vowel / LP Clean / LP Moog / LP Diode / LP K35 / HP Clean / HP Moog / HP Diode / HP K35 / BP Clean / BP Moog / BP Diode / BP K35 / BS Clean / BS Moog / BS Diode / BS K35
Off
Manual
Filter A
0 to 100 %
62
6
Filter B
0 to 100 %
22
7
Filter C
0 to 100 %
0
8
Vowel
i / y / u / e / oe close / o / schwa / eps / eps nasal / oe / oe nasal / open o / open o nasal / a / back a / back a nasal
i
Manual
Inertia
0 to 100 %
72
9
Presets
Subtle Flange
Classic Jet
Slow Sweep
Fast Sweep
Positive Feedback
Negative Feedback
Drum Flange
Guitar Flange
Synth Flange
Short Comb Sweep
Wide Jet
Metallic Sweep
AI Motion Add
AI Static Breakup
Extreme Jet
Flanger Safety Check
Gainer 5 controls · 5 automation targets
Changes gain, pan, and channel polarity for this rack.
Free rate / Gate / Attack / Release / Depth / Swing / Stereo shift / Mix / Output / Inertia / Gain / Panning / Offset / HP Cutoff / Low Gain / Mid Gain / Mid Freq / Mid Q / Hi Gain / 31 / 63 / 125 / 250 / 500 / 1k / 2k / 4k / 8k / 16k / Q / 60 / 12k / Azimuth L/R / Elevation / Distance / Width / Focus / Reflect / Room / Height Cue / Rear Cue / Air / Frequency / Drive / Harmonics / Color / Peak Guard / Air Limit / Amount / Carrier Note / Transpose / AM Bias / Stereo Phase / Low Cut / Anti-Alias / Tone / Bit Depth / TPDF Dither / Noise Shape / Clock Jitter / Reconstruction / Stereo Link / L Delay / R Delay / L Feedback / R Feedback / Send / Thresh / Hold / Floor / Feedback / Delay / Phase / Amplitude / Filter A / Filter B / Filter C / Ceiling / Bass Width / Bass Split / Decorrelation / Mono Safety / Balance / Damping / Stereo Spread / Resonance / Ripple / Low X / High X / Low Threshold / Mid Threshold / High Threshold / Low Ratio / Mid Ratio / High Ratio / Soft Knee
Manual
Amount
-100 to 100 %
0
4
Offset
-100 to 100 %
0
5
Min
0 to 100 %
0
6
Max
0 to 100 %
100
7
Smooth
0 to 1000 ms
80
8
Polarity
Unipolar / Bipolar
Unipolar
9
Inertia
0 to 100 %
72
A
Presets
Default
Preset 02
Preset 03
Preset 04
Preset 05
Preset 06
Preset 07
Preset 08
Preset 09
Preset 10
Preset 11
Preset 12
Preset 13
Preset 14
Preset 15
Preset 16
Line In 8 controls · 7 automation targets
Mixes one unique system capture input into this track. A physical input can be claimed only once per song.
Control
Range / choices
Default
Automation index
Input Source
No source
No source
Manual
Input Gain
-60 to 24 dB
0
0
Pan
-100 to 100 %
0
1
High Pass
5 to 500 Hz
20
2
Invert Phase
Off / On
false
3
Mix
0 to 100 %
100
4
Output
-24 to 12 dB
0
5
Inertia
0 to 100 %
72
6
Presets
Line In Default
Line In Transparent
Line In Gentle
Line In Warm
Line In Bright
Line In Deep
Line In Tight
Line In Wide
Line In Narrow
Line In Rhythmic
Line In AI Cleanup
Line In AI Restore
Line In Mix Polish
Line In Creative
Line In Strong
Line In Safety Check
LofiMat 11 controls · 11 automation targets
True sample-and-hold rate reduction with TPDF dither, bounded error-feedback noise shaping, jitter, and reconstruction filtering.
Control
Range / choices
Default
Automation index
Bit Depth
4 to 16 bits
12
0
Sample Rate
1000 to 48000 Hz
22050
1
TPDF Dither
0 to 100 %
24
2
Noise Shape
0 to 100 %
20
3
Clock Jitter
0 to 100 %
0
4
Reconstruction
0 to 100 %
35
5
Stereo Link
0 to 100 %
70
6
Character
Clean PCM / Vintage Sampler / Game Console
Clean PCM
7
Mix
0 to 100 %
30
8
Output
-18 to 6 dB
0
9
Inertia
0 to 100 %
72
A
Presets
Clean 16-Bit
Warm 12-Bit
Crunchy 8-Bit
Arcade 6-Bit
Console Rate
Sampler Rate
Telephone Rate
Soft Quantize
Noisy Tape
Digital Dust
Dark Lo-Fi
Bright Lo-Fi
AI Texture Add
AI Sterile Breakup
Retro Computer
Lo-Fi Safety Check
Maximizer 7 controls · 7 automation targets
Stereo-linked 4x FIR peak detection, genuine 6 ms lookahead plus detector delay, and smooth gain reduction without clipping. Place last; disable AGC when assessing loudness.
Control
Range / choices
Default
Automation index
Boost
0 to 24 dB
3
0
Threshold
-24 to 0 dB
-3
1
Peak Release
20 to 500 ms
100
2
Program Release
100 to 3000 ms
600
3
Ceiling
-6 to 0 dBTP
-1
4
Amount
0 to 100 %
100
5
Inertia
0 to 100 %
72
6
Presets
Clean presence
Gentle master
Peak protection
Streaming headroom
Vocal forward
Acoustic balance
Piano sustain
Bass stability
Drum presence
Percussion peaks
Electronic density
Rock bus
Quiet mix lift
Transient reserve
Parallel presence
Conservative ceiling
Maximizer: transparent peak control
Raise average level while controlling peaks with stereo-linked gain reduction, not clipping or saturation. A long 4x FIR detector anticipates peaks; a finite attack and program-dependent release preserve balance and avoid intentional harmonic coloration. Six milliseconds of lookahead plus 128 detector frames add about 8.67 ms at 48 kHz. A conservative 0.8 dB detector margin means the output can remain below the requested ceiling.
MLT 5.04: Actual knob view at 100% scale. All controls are automatable, including Inertia. Sixteen curated presets are available.
Place Maximizer last in the Master FX rack when protecting the complete mix. A track Maximizer cannot prevent overload caused by summing other tracks or by effects placed after it.
Disable AGC for comparison. Also disable the global limiter and avoid downstream gain boosts. The Maximizer does not change these global settings automatically.
Start with Clean presence or Gentle master. Boost adds input gain. Lowering Threshold adds explicit makeup gain: Boost + max(0, Ceiling - Threshold), in dB, at Amount 100%.
Ceiling sets the output peak target. Peak Release controls isolated transients; Program Release slows recovery on sustained limiting. Longer release can help bass stability. Excessive drive still reduces dynamics and can pump.
Amount blends the boosted and original signal BEFORE protection. Even Amount 0% retains ceiling protection; disable the device for actual bypass. Inertia smooths parameter movements.
The same graph is used in live playback, offline WAV, RP2S and sample-editor TFX. TFX compensates the processing delay, retains the selection length and supports Undo. This is not an RMS or LUFS target normalizer, a denoiser, or a guarantee of distortion-free output at unlimited gain. Existing source noise rises with gain. Monitor the output peaks when increasing gain.
Meta LFO 17 controls · 14 automation targets
Control-rate oscillator that modulates one installed DSP parameter on this track.
Control
Range / choices
Default
Automation index
Shape
Sine / Triangle / Square / Saw Up / Saw Down / Random / Custom
Free rate / Gate / Attack / Release / Depth / Swing / Stereo shift / Mix / Output / Inertia / Gain / Panning / Offset / HP Cutoff / Low Gain / Mid Gain / Mid Freq / Mid Q / Hi Gain / 31 / 63 / 125 / 250 / 500 / 1k / 2k / 4k / 8k / 16k / Q / 60 / 12k / Azimuth L/R / Elevation / Distance / Width / Focus / Reflect / Room / Height Cue / Rear Cue / Air / Frequency / Drive / Harmonics / Color / Peak Guard / Air Limit / Amount / Carrier Note / Transpose / AM Bias / Stereo Phase / Low Cut / Anti-Alias / Tone / Bit Depth / TPDF Dither / Noise Shape / Clock Jitter / Reconstruction / Stereo Link / L Delay / R Delay / L Feedback / R Feedback / Send / Thresh / Hold / Floor / Feedback / Delay / Phase / Amplitude / Filter A / Filter B / Filter C / Ceiling / Bass Width / Bass Split / Decorrelation / Mono Safety / Balance / Damping / Stereo Spread / Resonance / Ripple / Low X / High X / Low Threshold / Mid Threshold / High Threshold / Low Ratio / Mid Ratio / High Ratio / Soft Knee
Manual
Amount
-100 to 100 %
0
4
Offset
-100 to 100 %
0
5
Min
0 to 100 %
0
6
Max
0 to 100 %
100
7
Smooth
0 to 1000 ms
80
8
Polarity
Unipolar / Bipolar
Unipolar
9
Reset
0 to 255
0
A
One Shot
Off / On
false
B
Curve
Step / Linear / Smooth
Linear
C
Inertia
0 to 100 %
72
D
Presets
Default
Preset 02
Preset 03
Preset 04
Preset 05
Preset 06
Preset 07
Preset 08
Preset 09
Preset 10
Preset 11
Preset 12
Preset 13
Preset 14
Preset 15
Preset 16
Meta LFO: continuous musical modulation
The LFO now offers Custom in addition to sine, triangle, square, rising/falling saw and random. Use Curve to edit up to 64 points: click to add, drag to move, right-click to remove. Choose Step, Linear or Smooth interpolation. Smooth uses bounded easing between neighbouring points, not an exact reproduction of Renoise cubic curves.
Free rate is cycles per second. Line, Beat and Pattern rate is cycles per corresponding musical unit: 0.125 in Line mode means one cycle per eight lines. Synchronised phase progresses between rows from the scheduled audio-clock segments rather than jumping with the UI cursor. Row delays and changes of pattern length use the actual segment timing. The existing 20 Hz control-rate budget remains; this is not audio-rate modulation.
Reset restarts the cycle. Its byte parameter is automatable through Ixx/Jxx; writing the same Reset value again still retriggers. One Shot runs a cycle once and holds its final value until Reset. One Shot and interpolation are also automatable. Random holds one deterministic pseudo-random value per cycle and advances to a new value at each subsequent cycle.
Actual MLT 5.04 custom LFO curve editor with One Shot and smooth interpolation
Curves and settings are stored in .MLT. The LFO has fourteen addressable automation parameters, below the sixteen-parameter limit. Target routing retains MLT's forward-only rules. Free mode can evolve with an active audio clock while stopped; musical modes pause when song playback stops.
Meta Visualizer 11 controls · 11 automation targets
Meta device.
Control
Range / choices
Default
Automation index
Effect Type
BACKGROUND / 2D LAYER / 2D TRANSITION / 3D VOLUME / 3D TRANSITION / MATERIAL / ANIMATION / SCREEN FX
Three-head overlap-add reverse delay with click-free windowing, filtered feedback, stereo diffusion, and smooth time changes.
Control
Range / choices
Default
Automation index
Time
20 to 2000 ms
360
0
Feedback
0 to 92 %
32
1
Crossfade
2 to 250 ms
28
2
Low Cut
20 to 2000 Hz
120
3
High Cut
1000 to 20000 Hz
9000
4
Ping Pong
Off / On
false
5
Diffusion
0 to 100 %
22
6
Modulation
0 to 12 ms
1.2
7
Stereo
0 to 200 %
115
8
Quality
Eco / High
High
Manual
Mix
0 to 100 %
35
9
Output
-18 to 6 dB
0
A
Inertia
0 to 100 %
72
B
Presets
Reverse Delay Default
Reverse Delay Transparent
Reverse Delay Gentle
Reverse Delay Warm
Reverse Delay Bright
Reverse Delay Deep
Reverse Delay Tight
Reverse Delay Wide
Reverse Delay Narrow
Reverse Delay Rhythmic
Reverse Delay AI Cleanup
Reverse Delay AI Restore
Reverse Delay Mix Polish
Reverse Delay Creative
Reverse Delay Strong
Reverse Delay Safety Check
Revolver 5 controls · 5 automation targets
Mono/stereo impulse-response convolution. Load a WAV IR while stopped; IR audio is stored in MLT. Linear convolution reproduces cabinet or room responses, not nonlinear valve distortion.
Control
Range / choices
Default
Automation index
Pre-delay
0 to 500 ms
0
0
High Cut
200 to 20000 Hz
20000
1
Mix
0 to 100 %
100
2
Output
-24 to 12 dB
0
3
Inertia
0 to 100 %
72
4
Presets
Revolver Default
Revolver Transparent
Revolver Gentle
Revolver Warm
Revolver Bright
Revolver Deep
Revolver Tight
Revolver Wide
Revolver Narrow
Revolver Rhythmic
Revolver AI Cleanup
Revolver AI Restore
Revolver Mix Polish
Revolver Creative
Revolver Strong
Revolver Safety Check
Ringmod 13 controls · 12 automation targets
Band-limited stereo ring modulation with controllable AM bias, phase offset, sideband filtering, and equal-power parallel mixing.
Control
Range / choices
Default
Automation index
Oscillator
Sine / Triangle / Square / Saw
Sine
0
Carrier Note
0 to 119
36
1
Transpose
-36 to 36 st
0
2
Depth
0 to 100 %
100
3
AM Bias
0 to 100 %
0
4
Stereo Phase
0 to 180 deg
0
5
Low Cut
5 to 2000 Hz
20
6
Anti-Alias
1000 to 22000 Hz
16000
7
Tone
-100 to 100 %
0
8
Quality
Eco / Normal / High
High
Manual
Mix
0 to 100 %
50
9
Output
-18 to 6 dB
-3
A
Inertia
0 to 100 %
72
B
Presets
Clean Tremolo
Slow Amplitude Drift
Subtle Sidebands
Warm Bell
Glass Bell
Metallic Pulse
Robot Voice
Shortwave Radio
Bass Modulator
Octave Carrier
Fifth Carrier
Square Chopper
Triangle Motion
Sci-Fi Sweep
Broken Transmission
Extreme Ring
Router 3 controls · 2 automation targets
Routes this point in the track DSP chain to a later track and optionally mutes the source track post-FX output.
Revolver loads impulse responses for linear convolution. It can reproduce a captured linear response, not the full nonlinear behaviour of a driven valve amplifier. Autotune is intended for monophonic pitch material. Stereo enhancement and surround should be checked in mono when compatibility matters.
10 / Audio engine and analysis
MLT 5.04: Master spectrum in perspective waterfall mode. Click Spectrum or the analyser surface to cycle its six views.
Control family
Available choices / purpose
Output sample rate
Native, 22 kHz, 44 kHz, 48 kHz and 96 kHz settings. Actual device output and browser resampling remain part of the path.
In Audio / Perf, Mixer / Mastering, Amplify is a horizontal slider with all integer levels from x1 to x32 and a live value on its right. The default is x10. The rectangular thumb uses the same styling as MLT's other horizontal sliders. The first-launch master output volume is 100%; saved master-volume settings remain respected. Initial Global Volume in Song Levels is separate, and imported song-volume values are not overwritten.
The scale follows ft2-clone's floating-point output convention: the amplification multiplier is level / 32, multiplied by the independent master volume. Thus x10 denotes mixer level 10, not a tenfold gain applied to an already normalized output. Moving from x1 to x10 multiplies amplitude by ten. AGC can counteract level changes; disable it when comparing amplification. Parameter changes are smoothed in the output graph.
MLT saves the new level in native project output settings. The retired Off / +25% / +50% / +100% setting is not interpreted as a numbered FT2 level: projects without a saved new level start at x10. Legacy module formats do not store this application-output preference. Reference: ft2-clone, setAudioAmp().
Selected-track or master analysis includes stereo waves, spectrum, phase and loudness/peak displays. Spectrum modes include an outline, a filled curve, bars and a scrolling colour spectrogram. Pointer readouts expose frequency and level. Momentary, short-term and integrated loudness readings complement RMS and peak indicators.
Changing visual detail does not mean removing musical processing. The application separates audio scheduling from UI updates, uses bounded work and workers for costly tools, and uses asynchronous shader preparation in the Visual Lab. Lower visualization profiles reduce display work. High FFT precision, many DSPs, convolution, large samples and complex visuals still have real costs.
A lower requested sample rate may not dramatically change a source already limited in bandwidth. A browser can resample to the physical device rate. Likewise, 32-bit float mainly increases processing headroom and precision; it does not invent high-frequency information absent from a sample.
Spectrum: logarithmic, linear and waterfall
Click Spectrum or its analyser surface to cycle the unfilled curve, filled curve, bars, logarithmic spectrogram, linear SpectroTag view and perspective waterfall. The linear view uses a teal/amber gradient: image rows run from 0 Hz at the top to 22.05 kHz at the bottom, while time scrolls left and new audio enters on the right. Its upper frequency is limited by the actual hardware Nyquist frequency. The waterfall shows frequency, magnitude and recent history as receding traces; it is a bounded Canvas projection, not a separately compiled 3D scene. Hover for frequency and current-bin dB. Both alternatives share the existing master/track source selection and Audio/Perf refresh policy.
11 / MIDI files and hardware
Files
Whole-song MIDI import groups instruments and uses multi-note tracks, note delays, pitch and volume commands where the tracker representation permits. A bank provides sound; MIDI itself contains events, not sampled audio. Full-song MIDI export translates supported notes, timing and controllers, not the rendered sound of MLT DSPs.
Pattern-track MIDI IN imports the first applicable program/channel into the current track and pattern, truncating to that scope and expanding note columns where possible. It is available while stopped in Edit mode. MIDI OUT exports the selected pattern/track with a contextual filename.
Live input and output
Web MIDI input can audition or record notes, velocity and supported controller information. Instrument output settings include channel, program and pitch-bend range. Hardware availability and permissions depend on the browser and environment.
MIDI Learn
Right-click a supported slider, knob, transport control or Visual Lab randomizer to open mapping. Move a hardware control to identify it. Mapped controls pulse orange; unmapping restores their normal appearance. Supported mapping models include absolute controllers, relative encoders and button/pad behaviour, with applicable range and takeover settings.
Mappings use stable targets rather than screen coordinates. Project storage preserves configuration, but the receiving machine must reconnect its own MIDI ports. Controller resolution, relative-encoder conventions and browser support vary; test your hardware before a performance.
Build an audio-reactive scene from Background, 2D Layer, 2D Transition, 3D Volume, 3D Transition, Material, Animation and Screen FX. Each category has 16 choices. Enable categories independently, choose effects, randomize and save custom settings in the project.
Each effect exposes four mini-knob lanes with parameter-selection lists. A selected parameter is excluded from the other three lanes. Effects have their own preset variations and a Custom state. The current parameter pool includes shared timing/audio controls and, for selected newer effects, additional intrinsic parameters.
Assign audio from a selected track to visual controls, or use global randomization with the master source. Audio can influence movement, scale, rotation, speed and colour. Pattern commands and Meta Visualizer make those choices repeatable within a song.
Warm up effects before a performanceMLT 5.04: Visual Lab warm-up buttons below Randomize All
Below Randomize All in the Visual Lab panel, Warm up used inspects enabled slots, all stored patterns including extra effect columns, and every installed Meta Visualizer in the master and track racks. It follows Bxx jumps, Dxx row breaks, E6x loop branches, pattern repeats, the song restart position and marked-range returns while retaining visual focus and correlated 3D volume, transition, material and animation choices. Direct order starts and the current cursor provide additional entry points. It prioritizes combinations reached through these paths before textual fallback candidates. Random commands include every possible effect in their destination slot; DSP focus/value automation with Meta Visualizer admits all effect families. Analysis is limited to 4,096 states, a 1.5-second exploration window and 512 concrete 3D identities; preparation retains at most 1,024 programs per pass. A partial-combination notice means some random combinations or paths remain unprepared and may compile on demand. The tracker does not allocate an unlimited permutation cache.
Warm up all first displays a warning that preparation can take a long time. Choose Confirm to start or Cancel to leave the scene unchanged. The separate Cancel control remains available once preparation starts.
The warm-up status shows actual elapsed time from the start, followed by the estimated time remaining in parentheses: Elapsed 1:23 (remaining ~4:56). Elapsed time includes analysis and renderer initialization. Remaining time is an estimate, initially shown as estimating; after cancellation it is unavailable. The clock stops when preparation finishes or is cancelled.
Warm up all prepares every procedural effect variant and all 2D layer endpoint pairs, including effects not used by the current song. Both actions use the live renderer's shader cache without changing your scene, audio sources or pattern data. The visual animation pauses during preparation and resumes afterwards when the panel is active; the interface and Cancel remain available. Already prepared programs are reused. A progress bar reports ready programs and an estimated remaining time in minutes:seconds. Cancel stops additional submissions; a driver compilation already submitted cannot be interrupted.
Preparation is not limited to the ten-second startup budget. Keep the same browser tab and WebGL context open: compiled programs cannot be carried to another context or saved inside a project. You can resize the panel or switch editor views after preparation; the Visual Lab follows its available screen size without rebuilding unchanged render targets. Video capture keeps its explicitly selected dimensions. Switching to Mix preserves the selected DSP rack, changing a Pattern FX rack to Track DSP when needed. For very large randomized 3D spaces, individual effects are prepared against the selected scene rather than compiling millions of permutations. New combinations of volume, material, animation and transition may still need compilation; preparation reduces first-use work but cannot guarantee zero display stalls. A 1024-program safety budget includes ready and pending programs.
Persistent preparation cache. MLT saves prepared procedural shader sources in this browser profile, with a source-text budget of 8 MiB and at most 192 entries. A renderer-code fingerprint and individual SHA-256 checks discard outdated or damaged entries. Subsequent launches can reuse this preparation; disabled storage falls back to normal generation. Clear cache removes only these prepared sources, keeping your song, assets, Undo history and the current live GPU programs.
This is not a file of compiled GPU binaries. WebGL programs remain tied to their current context; the browser or Electron separately manages its own GPU cache. Browser, driver or hardware changes can require compilation again. The desktop edition preserves its per-user profile and stable local origin between launches. No machine-specific GPU cache is distributed with the installer or embedded in MLT projects. The startup logo remains visible for at least five seconds, with no more than ten seconds of optional startup preparation.
Desktop compiled-program reuse: Electron submits a reference link in a background Worker without waiting on unfinished GPU compiler work. After the matching display program completes asynchronously, a short reference-validation stage lets its native cache retain the compiled program for later sessions. Native program-memory and disk-cache capacities are each bounded at 32 MiB; this capacity is not preallocated and is separate from the source cache. The first preparation can still take time; later reuse depends on the unchanged shader, graphics backend, driver and retained application profile. The original effects and cancellation are preserved. A Worker failure falls back to parallel display preparation; the status reports if native-cache preparation was unavailable. The startup display keeps its five-to-ten-second policy. Clear cache above clears prepared sources, not Electron's native graphics cache.
Red effect names have no completed matching effect program in the current context. Their normal theme color returns after preparation completes. Restoring a source from disk alone does not mark an effect ready. Layer/volume combinations may still require an additional program even when their individual effects have been prepared. Some operating-system native option menus ignore custom text colors; the selected control also reports readiness through its text color.
When the parallel-compilation extension is missing, Firefox can use a background Worker with OffscreenCanvas. MLT submits the real program to the display context without immediately waiting for validation, flushes the request, and runs an identical reference compilation in the Worker before validating the display program. No shader or effect is replaced. The actual display-context program, not the Worker's private program, is retained for playback. Repeating preparation reuses it.
This Worker-assisted path is available to the two manual warm-up buttons; it does not extend automatic startup beyond ten seconds. Cancel terminates its Worker and releases unfinished display programs. Drivers differ: reference completion is not a standardized completion fence for another context. MLT measures submission, flushing and final validation, and pauses further preparation if one exceeds its 32 ms interface budget. A slow call already in the driver cannot be interrupted. Software WebGL, unavailable Worker WebGL or restrictive browser policies produce explicit feedback rather than silently reducing visual quality.