Release notes

Installing and opening

• RaveFoundry has a Windows installer. It installs into Program Files for everyone on the machine and adds a Start-menu entry. A newer version replaces the older one in place, and uninstalling leaves your songs, settings, presets, sound library and recovery files exactly where they are. It is not code-signed yet, so Windows shows an unknown-publisher warning when you run it.
• Double-click a RaveFoundry file and the right thing happens. Songs (.rfsong), song bundles (.rfbundle), backups (.rfbackup), content packs (.rfpack) and Rave Shows (.rfshow) all open in RaveFoundry, and each does its own job rather than merely starting the app: a bundle unpacks and opens, a backup is checked and staged for restore, a pack opens the Sound Library ready to preview, and a show opens the song it belongs to.
• The installer is tested by being used. Every release build is packaged, installed and started from that package before it can ship. An installed copy and a development build also keep their own data folders, so one can never disturb the other's songs and settings.

What it says about itself

• The two Vowel dials describe the filter that ships. It is three resonant peaks moved through five formant settings, a filter colour rather than a spoken vowel, and the labels now say so. The effect itself is untouched.
• The cheat sheet and the manual say what the keyboard really reaches. Section headers are tab stops, dials are set with the mouse or Alt+wheel, and the EQ number boxes take typed values. Both are written from what the app does.
• Screen readers get the real thing. Grid cells, track rows, notes, meters, drop chips, the library and the tour cards each report their name and their state, and a dial reports a settable value in its own units. Faders, custom controls and every drop-down announce their real names too.

Which version wrote this file

• A song records the RaveFoundry that saved it. Open one written by a newer RaveFoundry and it is refused, with the message naming both versions. Open a song from an older RaveFoundry and the status line says once that saving it is a one-way trip.
• A Rave Show from a newer RaveFoundry is left alone. It used to be reported as missing, or overwritten. Now it is named and left exactly as it is, and cue kinds this version does not know are kept rather than dropped.

In the sequencer

• Cell sizes are Medium and Large. Small has gone from the CELL strip, and a song saved at Small opens at Medium.

Your work, protected

• A crash costs you seconds, not minutes. An unexpected error now copies your unsaved song to a recovery file before the error dialog appears, instead of leaving you with whatever the last background autosave caught. The dialog tells you whether the copy was written, or says to save the song yourself if it could not be. The next launch offers that copy back and says the app closed unexpectedly, so you can tell it apart from an ordinary autosave.
• Opening a song offers back the unsaved changes sitting beside it. The launch-time recovery offer only ever looked at your eight most recent songs, so a song opened from anywhere else had its recovery copy quietly overwritten the first time you saved. Now the open asks first, however long ago that work was: accept to pick it up, decline to open the saved version. A recovery copy that will not load is kept rather than deleted, and the saved song opens instead.
• Replacing a song that still has unrecovered changes asks first. Saving over an existing song file used to delete that song's recovery copy as routine bookkeeping: work from an earlier session that had never been offered back to anyone. RaveFoundry now names the song and warns what replacing it discards: decline and nothing at all is written. Importing a song bundle and choosing Replace mine asks the same question, and so does saving where the song itself is gone but its recovery copy is still sitting there. An ordinary Save, and a save onto a song with nothing unrecovered beside it, are unaffected.
• A bundle you share carries all of its audio, or says what is missing. If a recording the song uses has been moved, renamed or left on a drive that is not plugged in, Export stops before it writes anything and names the clips or banks it cannot find, so you can put them back. If they are gone for good you can still choose Export anyway, and the bundle then tells whoever opens it which pieces the sender did not have. Saving a copy of a song into a different folder warns you the same way, instead of writing a copy that looks complete and quietly is not.
• Your saved presets can no longer be wiped out by a file RaveFoundry cannot read. The preset bank now carries a name and a version: one written by a newer RaveFoundry is left completely untouched and named on screen instead of being replaced by an empty one, a damaged bank is set aside under a dated name so the old file is still there to recover, and a preset for a track type this version does not know is kept and written back out rather than dropped. While a bank is being left untouched, every preset you save says so as you save it, so you are never told a preset was saved that will not be there next time.

Help where you are

• Help ▸ Release notes. What changed in this version and the ones before it, read straight from the app. No browser, no connection.
• Ctrl+F1 opens a one-page sequencer legend. Every mark a step can carry and every mark the grid draws around it: the swung gridlines, the two playheads, group headers, pinned rows and the transitions row, all on one page, from any window. Also on Help ▸ Sequencer Legend.

What it says about itself

• RaveFoundry now states what it needs from your machine. Getting started ▸ What RaveFoundry needs from your system opens with a minimum-and-recommended table: Windows 11 64-bit, 4 cores, 8 GB RAM, a 1920 × 1080 display and 500 MB of free disk at minimum; 8 cores or more, 16 GB RAM and a 2K or 4K display recommended. It says the app itself typically uses under 500 MB of that memory, and spells out the case a spec sheet hides: a 1920 × 1080 panel set to 125% or 150% scaling gives the app less room than the layout is designed at, even though the panel is not below the requirement.
• The manual documents the struck-body voice. Forge Body gains its own row in the track-types table, covering the geometries, what Ring Time, Damp Tilt, Hardness and Strike Pos do, and that the tail is the note. The chapter's type counts now agree with the 26 the app ships, and the Audio track no longer advertises a bake that does not exist. The sound chapters and the generated parameter reference describe the sounds in the app's own words.
• The document RaveFoundry's claims are written from is honest again. It had gone thirteen releases without a check, and some of its numbers had quietly gone false: the song file version, the smallest window size, the key that opens the keyboard cheat sheet. All are corrected against the code, and Sound Match, the Master and Low-End Labs, Sonic Contracts, Constrained Variation, song bundles, backup and restore and the v0.27 shell are described for the first time, each with what it will not do.
• Getting started matches the app again. Its audio paragraph had said for several releases that there is no output device to choose and that RaveFoundry only plays and never records. It now names the picker in Settings ▸ Audio, says that a device you chose which later goes missing falls back to the Windows default and tells you so, and points at the input picker and the recording chapter beside it. The First Song tutorial's bass and pad steps name two presets the picker really carries: Acid Squelch and Warm Strings.

Plainer names

• The bundled templates and demos carry generic track names. Four style templates and four demo songs now label a track with what it is: Techno Kick, Acid Line, Sub Kick, Big Clap, Sub Dive, E-Piano, Ensemble Bed and Warm Strings. Only the labels moved: song titles, file names, presets and every parameter are untouched, and each song renders exactly as it did before. The controls follow. The Percussion Model dial's last three recipes read Bridged Rim, Sine Clave and Broad Cabasa, each named for the circuit that makes it, the hi-hat's choke switch is now Choke Mode with its two positions unchanged, and the hover text says what a control does. No sound, range, default or stored value moves, and Sound Match describes a match in the same plain terms.
• The shipped program no longer carries the developer's folder path or the repository address inside it. Nothing about how RaveFoundry runs changes; the files you install simply stop naming the machine and the private code repository they were built from.

Your work, protected

• Backup and restore carry everything now. A saved effect-rack chain is no longer left out of a .rfbackup because of the way you named it, and a large sound library backs up instead of failing. The file written is unchanged. And restoring a backup whose settings can't be read keeps this machine's preferences as they are, rather than resetting your zoom and window placement to near-nothing.
• Song bundles ask before they act. Importing one now raises the unsaved-changes question before it unpacks anything, so Cancel really does change nothing. It used to leave the bundle's files already written, and your own song already gone if you had chosen to replace it. It also asks what to do about a leftover audio folder or Rave Show of the same name, not only a song of that name. And exporting a bundle with unsaved changes now offers Save, Continue or Cancel: a bundle is built from the song on disk, so sharing one mid-edit used to hand over a song missing everything since your last save.
• A recovery copy is never tidied away before you have been offered it. When a crash strands several unsaved songs, the ones you have not been asked about wait for a later launch instead of being deleted. A recovery copy left behind by an older version of RaveFoundry is offered back to you again, instead of sitting there unseen.
• Opening a song while RaveFoundry is busy leaves your work alone. Double-click a song file while an export is writing, or while a dialog is open, and the running window simply comes forward. Open it again once the app is idle. Opening one in the moment RaveFoundry is closing now works too: the new launch waits for the old copy to go and then starts up itself, and tells you if it never does. And restarting to apply a new accent colour can no longer leave you with two copies open.
• Two exported files can no longer land on the same name. Queue rows that share a format and bitrate but differ in sample rate, tail or bar range each keep their own file, so one can never quietly overwrite another.

Playing and monitoring

• The automatic waveform refresh keeps to itself. Pausing and then waiting a couple of seconds no longer loses your place. Play resumes exactly where you left off, playhead marker and all. An edit made while the strip is redrawing is no longer thrown away: the strip stays dimmed and redraws itself when the render finishes. And with live MIDI input armed it stands down altogether, instead of taking the sound device for a moment after every knob tweak and cutting your monitoring with it. Refresh now still re-renders whenever you ask for it.
• The sound device changes hands cleanly. Turning the live MIDI monitor off just as playback starts no longer leaves the transport frozen and silent; closing RaveFoundry a moment after arming no longer holds the window open for a couple of seconds; a buffer size picked in Preferences always takes hold, including in the seconds after a pause where it could be quietly discarded; and losing your device while paused now says it will be re-opened on the next Play, instead of reporting that playback stopped and asking you to press play to retry.
• Space works inside a popped-out panel. A focused button presses and a focused drop-down opens, instead of the transport stealing the key. Space still starts and stops playback when nothing that needs it has focus.

Recording and arming

• The MIDI Record window keeps up with your session. Quantize last pass lights up as soon as a pass is recorded, instead of staying greyed out with no way to reach it, and greys out again when you start or open another song, where it used to stay lit over the recording you had just left and stamp that song's notes into the new one. Deleting, reordering or reloading tracks updates the Record button and the armed lines straight away, so Record no longer looks available once the only armed track is gone.
• Recording in the Audio Lab tells you when it stops on its own. Losing the microphone mid-take, or reaching the ten-minute length limit, used to leave the panel saying it was still recording, for good. The take is now finished and saved, and the panel says which happened.
• Closing the Controller Mapping window mid-learn cancels the learn. It used to stay waiting with its prompt gone, so the next knob or key you touched was bound and saved without a word. Your existing mappings, and MIDI input itself, are untouched by the close exactly as before.

Buttons, labels and help

• Buttons tell the truth about what they can do. Show in Explorer, Export song bundle and the Rave Show Save/Load buttons grey out the moment you start a New song, a New from style or a demo, and light up again once a recovered song is restored. File ▸ Export MIDI follows as you add and remove tracks. The export dialog's Add to queue greys out, with a tooltip saying why, once the queue holds its eight rows, instead of hiding the ninth click's refusal behind the dialog. And a failure while preparing a render, export, freeze, stems bounce or bake no longer leaves the app convinced an export is still running, with File, Export and Undo stuck.
• Dials and drop-downs read what they mean. The kick's Click Type calls its fourth setting Beater, not "Tick". The saturation-character dials on the master multiband and the Creative Bus read Tape / Tube / Fold / Clip, the same words the per-track saturation dial has always used, instead of a bare 0–3. A Bass, Lead or Pad track's LFO Target now offers PWM alongside Cutoff, Pitch and Amp: it was already reachable from the dial, a preset or a step lock, so merely selecting such a track could quietly send its target back to Cutoff. And OCT lists only the octaves the selected track can actually use, instead of offering entries that snapped back the moment you picked one.
• The manual caught up. Help ▸ Keyboard Shortcuts names its real key, Shift+F1; the troubleshooting page names your backup and autosave files the way the app really writes them; and the autosave page says what is true: every never-saved song has its own recovery slot, and a crash that stranded several offers each of them back.

Share a whole song

• File ▸ Export song bundle… writes a song, the audio it plays and its Rave Show into one shareable .rfbundle. It carries every sample the song uses: recorded takes, imported clips, revert-to-source audio and resynth wavetables, including any that live only inside a Named Snapshot, so a snapshot you kept never loses its sound. Licensed factory sounds travel inside your own song without complaint.
• File ▸ Import song bundle… checks every file inside against a recorded fingerprint before it writes a single byte. A damaged or tampered bundle is refused whole and nothing lands on your disk. It then unpacks the song and its audio folder beside the bundle and opens it. When a song of that name is already there it asks what to do: keep what you have, keep both (the incoming song arrives under a new name, with its own audio folder), or replace yours with the bundle's. It never silently overwrites a song.
• A bundle is built from the song on disk, so Export waits until the song has been saved. To hand someone the raw files instead, File ▸ Show song in Explorer selects the .rfsong and its .rfassets folder together.

Backup and restore your app data

• File ▸ Backup app data… writes everything you author outside a song (your settings, saved track presets, effect-rack chains and your whole sound library) into one .rfbackup file, so a reinstall or a new PC does not start from nothing. Your songs are not in it: they live where you saved them, each beside its own .rfassets folder (to carry one song and its samples, share a song bundle).
• File ▸ Restore app data… brings a .rfbackup back, and applies it at the next launch, since settings and presets load once at startup. Restore checks the file, explains it will apply next time, and offers to quit now, keep working, or cancel so nothing changes; the first window after says what it restored and when the backup was taken. A damaged file, or one from a newer version of RaveFoundry, is refused by name before it touches your data.
• A restore is reversible and machine-aware. The stores it replaces are moved aside, not deleted, into a one-step undo slot, so restoring your previous backup puts things back. This PC's audio device, MIDI-device hint, remembered folders and a custom library path that does not exist here stay this PC's; everything else, your window layout included, comes across.

Nothing slips away unsaved

• Opening or starting a song now asks about unsaved changes, not just closing the app does. New, New from style, Open, Open Recent and Open Demo, and opening an imported song bundle, all raise the same Save / Discard / Cancel prompt you already get when you close with unsaved edits. Cancel backs out and changes nothing; Save runs the ordinary Save.
• A Windows sign-out or a restart to install updates no longer loses your last edits. As the session ends, RaveFoundry writes your settings and, for a song with unsaved changes, its recovery snapshot, so the next launch offers the song back, exactly as it does after a normal close.

Diagnostics and error reporting

• Every handled error now leaves a trace. The save, load, preset, pack, MIDI, export and clipboard messages each write the full details to the rotating crash log as well as showing them, and gain one line naming the file. What you see is otherwise unchanged, and nothing is sent anywhere.
• A diagnostics group under Help. Help ▸ Open logs folder and Help ▸ Open app data folder open those folders in Explorer. The app-data one is also the simplest way to make a manual backup, by copying the folder. Help ▸ Copy diagnostics puts your version, the Windows build, the runtime, your audio devices and the newest crash log's tail on the clipboard as plain text, ready to paste into a report.
• The crash dialog gained a Show in Explorer button for the log file it already named as text, so you can open the folder without copying the path by hand.

Help that lands where you asked

• Press F1 on any window and the manual opens at that window's chapter, not at the top of the page, and the ? button in a window's title bar lands the same place.
• A coverage gate keeps the manual honest. The build now fails if a window has no help chapter, or if a menu row is named nowhere in the manual, so a new surface cannot ship undocumented.

One RaveFoundry at a time

• A second launch no longer opens a rival window. If RaveFoundry is already open and you start it again (a second double-click, or opening a .rfsong from Explorer) the second launch hands its song to the copy already running and brings that window forward, then bows out without touching your settings or recovery data. When the running song has unsaved changes you get the usual Save / Discard / Cancel prompt before the handed-over song replaces it.

Your files, properly named

• Songs save as .rfsong. The Save dialog fills in the right extension by itself now, the audio folder beside a song is .rfassets, and the export sidecar is .rfcontract. Every song you already have keeps opening exactly as before. Old names load forever.
• Save As remembers where your songs live, and a fresh install starts you in Documents\RaveFoundry\Songs instead of wherever Windows last pointed.
• File ▸ Show song in Explorer jumps straight to a saved song and its files.

Pick your audio devices

• Output and input pickers live in the new File ▸ Settings ▸ Audio tab, beside the buffer size. "(Windows default)" stays the default. If your chosen device isn't plugged in, RaveFoundry plays on the default, says so, and keeps your choice for when it returns.
• A paused session follows a default-device change. Pause, switch the Windows default, press Play, and the sound comes out of the right device now.

Meters that agree with their faders

• Meter fills are on the decibel scale, matching last release's console-style faders. A −6 dB fader beside a −6 dB signal finally looks like it, peak-hold and the hot/silence colours included, and screen readers hear the reading in dB.

The words caught up

• "A/B Versions." The whole-song A/B feature stops sharing a word with the MIDI take bin. The transport toggle reads Versions, the button Capture version. Named Snapshots and MIDI Takes keep their names.
• Settings, under File. Preferences moved out of the Tools menu to File ▸ Settings, where an app-settings menu belongs.
• A real license. Help ▸ License opens the terms: your music is yours, and the factory sounds are usable in your works royalty-free. Help ▸ Third-party notices opens the component credits, and the About window carries the short version.

Honesty, litter, and recovery

• Crash recovery offers every stranded song, not just the newest, and cleans up the stale autosave files it used to leave beside your songs forever.
• Temp litter sweeps itself after a crash, and the waveform cache deletes leftovers from older versions.
• The app tells Windows who it is. The program is RaveFoundry.exe from EverUplift LLC, so Task Manager and file properties finally match the product.

Arm and record, for real

• Every MIDI track has a record-arm. An R joins mute and solo on the mixer strip: red when armed, absent (with the reason) on audio tracks, which record through the microphone instead. Arms are for the session: a saved song never re-opens armed.
• Tools ▸ MIDI Record. The whole recording surface in one window: how many tracks are armed, which track each keyboard channel lands on, replace or layer, quantize, sustain, punch-in/out, and the same count-in the transport uses.
• It refuses honestly. Loop-record greys the punch controls and says why: the loop window wins. Pattern mode refuses a record start outright instead of quietly growing your song.

Faders that drag like a console

• Every fader now moves in decibels. Unity sits at three-quarters of the throw on every fader, track, group, MASTER and Rave Mode alike, with a uniform working range down to silence, so the part of the throw you actually mix in stopped being crammed at the bottom. Your saved mixes are untouched; only the drag changed.
• The wheel speaks dB too: half a decibel per notch, a tenth on fine.

The rack learns to split properly

• Band split and dry/wet are editable at last. A split rack draws its devices banded under branch headers; pick Parallel, Band split or Dry/Wet, add and remove branches, move devices between them, and set per-branch gain, mix and ascending crossover frequencies exactly where each mode really uses them. Group racks get the same editor.
• A rack that can't honour its split says so. Removing a device used to leave it silently playing as a plain chain. Now it declares that and points at the fix.

Sound Match fades, and previews that stop lying

• Match previews stop cutting sounds short. A matched sound that rang past the reference sample's length was being chopped mid-tail in the preview, while the applied track played fine. Previews now get room to ring and land on an exact zero.
• Match fade in / Match fade out. Fade the matched sound over seconds, written into the fit's own envelope. What you hear in the preview is what Add, Clone and Save commit. Sounds that can't fade say so plainly (a struck drum's shape is the hit).

The library, the macros, and the honest details

• Track presets sort and favourite, samples gain a labelled Source filter, and the tag row becomes yours alone. Pack import now names exactly which items collide.
• Macros grew up. Each target row shows the live value it sends, and a + Target (learn) arms the card so the next dial you touch becomes the target. Right-click on a dial stays exactly what it was: type an exact value.
• Automation lanes draw their own shapes. Curved segments used to sound curved and draw straight; they draw the truth now, with a shape picker to match.
• Small honesty: a group-fader drag is one undo step, applying a designed sound no longer mislabels the track's preset, and demo track names went generic across the bundled songs.

Groups get a rack of their own

• Every group bus now has an editable effects rack. Open it from the group's gutter menu or from its new mixer strip and you get the same editor your tracks use, with the same twenty-two devices added and reordered the same way, working on the whole summed bus. One compressor across the drums, instead of five.
• And you hear it while you work. Group racks are new here, and a split was closed on the way in: a group's rack would have rendered into your export while staying silent in playback. Live and the bounce agree, and a standing check keeps them agreeing.
• Save a rack as a chain and use it again. Any rack, a track's or a group's, can be saved into your library, and a new Rack Presets tab in the Sound Library lets you browse, apply, rename and delete them. They live on your own machine; the file dialogs are still there for carrying a chain to another computer.

The mixer's missing half

• Your groups appear as strips beside MASTER, with a fader, mute, solo, the two group sends and a button straight through to the group's rack. Those sends were saved and rendered all along; this is the first release where you can move them.
• Every fader reads in decibels: a figure under each cap at rest, and the right-click entry now takes dB directly: type -6, or -inf for silence.
• All four per-track sends fold behind one control (delay, reverb, creative bus and fractal echo) and the strips fold together, so the faders stay lined up.
• Each strip's menu jumps to that track's EQ or its rack, scrolling the panel into view even when the inspector column is long.

An EQ that names its shapes

• The five gainless band types stopped sharing one dot. Low-pass and high-pass nodes now draw their own response in miniature, plateau, knee and slope, while notch dives, band-pass peaks and all-pass wears a diamond. The readout leads with the shape, so you can see what a band is doing without selecting it.
• Every shape explains itself. The type picker carries a plain sentence for each, including why a gainless band greys out its dB field.
• Two new ways in: Track EQ… in the Edit menu beside the pitch grid, and Master EQ under Tools.
• EQ curves travel. Save a track's curve, or the master's, to a file, load it onto either, or copy one straight from track to track. One undo, either way.

Leaner songs, fuller exports

• Songs take about a quarter of the disk they used to. Every step in a song file now writes only what makes it different instead of restating every default. The thirteen bundled demos went from 16.24 MB to 3.63 MB, and they play back exactly as they did before, note for note.
• WAV exports carry their tags too: title, artist and BPM, finishing the set OGG and MP3 started last release.
• The track-name column got its room back at the largest grid cell size.

The Audio Lab, with a way back

• Detect, drag and delete are all undoable. Re-running slice detection used to quietly replace every marker you'd moved by hand. One undo now takes it back, and dragging a marker is one undo, not one per pixel.
• "Revert to loaded" is a floor under the whole session. Three edits deep, past a spent undo, one click gives you back the audio you started with, and it still remembers the clip it came from, so applying it back knows where home is.
• The selection readout warns you before you send. It tells you when a selection is longer than Sound Match will take, right beside the button that would hit the limit.
• The send row reads by destination: into the song, into the library, out of the app, or across to another tool.

An export you can call off

• Cancel works, Escape included, and the disk is left exactly as it was. Every format now writes beside its target and moves into place only on a clean finish, so an abandoned or failed bounce can no longer leave a short file sitting on top of the good one you already had.
• The dialog remembers what you chose (all nine settings and your export folder) and seeds the file name from the song's own name.
• The export reports itself. On success the dialog stays open and shows where the file went, what the quality checks found and which export of this song this was, with buttons to open the folder or the report. A written QC report now lands beside every export. A labelled progress bar says which phase the render is in, and the range line carries a duration and size estimate.
• Several files, one render. Queue the deliverables you want, WAV 24-bit and MP3 320 say, and export once: rows that agree on everything the renderer hears share a single render, and one Save dialog names them all.
• Your exports carry their name. OGG and MP3 files go out tagged with title, artist and BPM (artist is a new preference: leave it empty and it's simply left off).

Variation that plays fair

• Hearing a candidate is free. Every candidate row has a ▶ that plays a short window of it. No take spent, nothing changed. The A/B slots still work for comparing mid-loop, and when keeping one does push older takes out, the window now says how many.
• You can see what's locked. Checklists beside the phrase box for tracks, for sound types and, new here, for whole dimensions: pitches, rhythm, velocity, articulation. When something is skipped, the refusal names the real reason ("busier skipped, rhythm locked") instead of blaming the track.
• Four moves that had no way in now have one: Backwards, Same groove new notes, Mirror the melody and re-articulation, on the preset row and in the plain-language phrasing.
• The seed saves with the song, so the same song, the same request and the same seed still give the same answers after a restart, and "More like this" generates around a candidate you liked.
• The preview stopped fibbing. Two-bar grooves draw all their cells, and a new line prices an Apply before it happens: "0 placed · 4 cleared · 4 rewritten · 120 unchanged".

Around the studio

• Solo is back on the track rows: the small S beside mute, the same state as the mixer's, live while the song plays.
• The View menu is flat. The panel toggles and their reset sit directly under View, and the Mixer opens from Tools now, alongside every other tool window.
• Tools is organized by what you're doing: make a sound · shape the song · mix & measure · play & record, alphabetical within each group.
• Every song and template in the box is on the current file format: the thirteen demos and the ten starter templates, moved through the app's own loader, with the handful of sounds that had drifted onto retired settings refitted to the ones their files always described.

One preview at a time

• The whole app now plays one preview at a time. Every ▶ button (the Library, Sound Match, the Audio Lab, the Sound Designer, MIDI Takes and MIDI Import) follows the same rules: starting a second preview stops the first, and a second click on the same button stops it instead of restarting it.
• Previews never play over the song. Starting a preview while the song plays is politely refused, and pressing play on the song stops whatever was auditioning. No more two things at once.
• Closing a window stops its previews. Including the Sound Designer's audition, which previously couldn't be stopped at all.

Measurement you can trust

• Master Lab readings admit when they're stale. Edit the song after an analysis and the panel says so, and names the kind of edit, instead of showing confident numbers for a song that no longer exists.
• A running analysis can be cancelled, a progress line says which phase it's in, and the report exports to a file. Songs measure up to about 30 minutes now, and when one runs longer the reading says exactly what it covered. Reference tracks you've compared against before are one click away again.
• The Low-End Lab opens on your song. Song mode is now the first tab, with the kick-and-bass reading front and centre, and a new opt-in duck measure reads how much your bass actually gives way under the kick, from real renders, with the cost stated before you spend it. Each tab exports its own reading.

Sound Match keeps your work

• Your sample and ranking survive a restart. A multi-minute fit used to die with the session; the panel now brings your sample and a summary of the last ranking back when it reopens.
• Refit just one voice. Any ranked row can re-run the search on its type alone, and find similar asks your Sound Library which of your own presets already lands close to a match's settings.
• Send audio straight to Sound Match from the Audio Lab, the whole buffer or just your selection. The panel now states plainly which voices it can search and which kinds of sound it can't answer for, rather than scoring them as poor matches.

Workflow

• The whole-song strip zooms with the mouse wheel, through sensible steps, and your zoom is remembered. It appears instantly when you reopen a song. If the song changed since, it asks before showing yesterday's shape, and says why it thinks the picture is old.
• Right-drag a bar range on the strip to send exactly those bars to the Audio Lab.
• MIDI Takes and MIDI Input link to each other, so the two halves of the same workflow no longer make you find the other in the menus.
• One spelling everywhere: Analyze. The app spoke two dialects; buttons, panels and the manual now agree.

Readable at every zoom

• Menus, dropdowns, right-click menus, tooltips and pop-ups now follow View ▸ Zoom. They used to stay tiny while the app around them grew; now the whole surface zooms together, up to the same 150% ceiling.
• The smallest text in the app got bigger. The track-group labels at the Small cell size were below any published floor and now match the grid's other smallest marks.
• Windows can't be shrunk into clipping their own controls at high zoom. Minimum sizes now understand the zoom you've chosen.

Help that knows where you are

• F1 opens this window's manual chapter in every window, from Sound Match to Preferences. Shift+F1 opens the keyboard cheat sheet from anywhere, and a new ? button sits in every title bar beside the ✕.
• The cheat sheet grew up. It now states the universal rules (Escape closes any tool window, F1 = this window's chapter) and covers the tool windows' own keys; the manual gained sections for the Preferences window and the app's dialogs.

Text you can trust

• Errors are legibly red. The status line's error colour sat just under the recognised contrast bar on its background. It now clears it, and errors, warnings and confirmations each read at full contrast.
• Cut-off text always has a tooltip. Sixty places could shorten a long name with an ellipsis and give you no way to read the rest; every one now shows the full text on hover, and a standing check keeps it that way.
• Tabs wear the theme. Four windows' tab strips were rendering in stock white-on-grey; they now match the app, with a clear accent edge on the selected tab.

Small things that respect your hands

• Every window puts focus somewhere sensible when it opens (the first field, the primary button, the editor surface) so Tab and Enter work immediately.
• Rave Mode's pads show a visible focus ring for keyboard players, and the rack's modulation rows re-flow instead of clipping when the panel is narrow.

One look, every window

• Every tool window wears the same title bar: the tool's colour dot, its real title, and a clean edge that dims when the window loses focus, so the active window is always obvious at a glance.
• Tool windows gained minimize and maximize. Sound Match, the EQs, the Labs, the Library, Routing and the rest were close-only; they now carry the full set, and maximize respects your taskbar on every monitor.
• Escape closes any tool window, every one of them, the same way. Windows with a save box open close that first, exactly as you'd expect.

Your windows remember

• Size, position and maximized state persist per window. Open the EQ where you left it, at the size you left it, next session.
• Restores are checked against the desk you actually have. A remembered spot on a monitor that is gone recenters instead of opening off-screen, and a window can never come back too small to use.
• Minimum sizes now understand interface zoom. At 150% zoom, windows can no longer be shrunk to the point of clipping their own controls.

Clearer messages

• Errors are red, warnings are orange, confirmations stay quiet. A failed export no longer looks identical to a saved file, and the warning lines that used to stack up folded into one place that shows only what is active.
• A recent-history flyout on the status line keeps the last sixteen messages with timestamps, so an outcome that was overwritten before you read it is no longer lost.
• Every message box is now the app's own. The system-grey prompts are gone, replaced by one dialog in the house look with the exact keyboard behaviour you already know: Enter confirms, Escape cancels where cancel exists.
• Tool windows report in their own status bars. Pattern Variation's results now appear in the Variation window itself, not on the main window behind it.

Preferences, and fixes

• Tools ▸ Preferences: a simple two-tab home for the settings that had none: the preset-apply behaviour switch, the MIDI input backend, and one-click resets for the tours, the remembered paste choice, and all tool-window placements.
• Pitch Grid's legend no longer overflows its window, a wrap that had been silently broken since it shipped, at every zoom level.
• The window-placement hardening from v0.26 now protects every window, not just the main one, and a damaged remembered position degrades to a clean recenter instead of an error.

Sound Match puts the fitted sound on your track

• Clone Track with Match. The third way out of a fit: a new track of the matched type carrying the fitted sound and your selected track's pattern: every step, its locks, its place in the mix. The original is never touched, and the whole thing is one undo.
• Preview on Track. Hear the swap in the mix before you keep it: the clone plays, the original mutes, and cancelling puts everything back exactly, even the unsaved-changes dot. Saving, exporting or loading while a preview stands cancels it safely first, so a preview can never leak into a saved song.
• The fit tells you how long is left, with a live estimate that tracks your machine's actual pace as it works, rounded up, never falsely precise.
• And it searches more of what the synths can already do. Sixteen dials across ten voice types joined the search: filter shapes on Lead and Pad, drive on the FM voices, the organ's drawbar taper, pluck damping and palm mute, the struck body's modal controls. All existed and were audible, and the fit simply never tried them before.

Playing dynamics reach the voices that ignored them

• Chord answers velocity. The curve, velocity-to-brightness and velocity-to-decay controls the other synth voices already had. A harder-played chord now opens up and tightens instead of only getting louder.
• FM Keys gains Vel→Index: velocity into the FM brightness itself, the defining touch of the electric-piano family, and it only ever softens below full velocity so the voice's ceiling never moves.
• The struck body gains a Damper. Release the key and the ForgeBody's ring dies faster, like a hand on a bell, up to ten times faster at the full dial. Every factory preset ships with it off, sounding exactly as before.
• Chord gains the bright-decay pair, upper partials that fade faster than the fundamental, the struck-string envelope. The kick family exposes its click length and beater width, constants your hands could never reach before.

Sonic Contracts: the song remembers what you approved

• Approve an export as the reference take, and every later export is measured against it: loudness, loudness range, true peak, mono compatibility, the limiter's work, and a fingerprint of the render. Warn-only: it never blocks and never changes your sound.
• It measures the real deliverable, riding the export itself, and where two takes honestly cannot be compared (a different sample rate, a different export setting) the report says why, per figure, instead of faking a verdict.

A new reverb: the Warehouse

• The rack's twenty-second device and its first true feedback-delay-network space. Eight delay lines, one absorptive filter per line, which is what lets its Decay be a real time in seconds (0.3 to 20) and its HF Decay lose the treble first, the way real rooms do. A diffused early field blends against the tail, Size moves the walls without changing the decay, and at Mix 0 it is a perfect bypass.
• It earned its place by ear. The first listening test failed: the tail sat too low and the early field coloured the source. It was rebuilt from that verdict and re-judged: ten of twelve rows good, none poor. Very short decays lean on the early field by design; the manual says so.

The low-end advisor

• Tools ▸ Low-End Lab can now read your song. Pick the kick and the bass (it guesses from the track types) and it reports where they fight: which low band they share, whose energy owns it, how the bass sits in time against the kick, and how much of the master limiter's low-band work each one causes. Every suggestion names the measurement behind it, and it only ever tells. It changes nothing.
• The surgical EQ gains an All-Pass band, the alignment tool the advisor's advice points at. It turns phase at its centre frequency without touching level. Place it yourself; no existing song changes.

Honesty, safety, and windows that stay on screen

• A song from a newer RaveFoundry is refused with a clear message instead of silently losing the track the older version didn't understand.
• The main window can no longer strand its menu off-screen. Restored positions are pulled back onto the desktop, a maximized window fits its monitor exactly, and View ▸ Reset window position is the escape hatch. On displays smaller than the design size the layout gets denser instead of hanging off the edge.
• The Vocal track's eleven modulation dials became four labelled groups (Chorus, Flanger, Phaser, Freq Shift) with not one sound changed.
• A rare crash on very high notes in the Wavetable voice is fixed, and live rack edits now take effect as one clean transaction with any refusal explained instead of swallowed.
• One new dial did not survive its listening test. A pluck release-noise was built, judged to make the sound less clean, and removed before shipping. Building, listening and removing is the process working, so it is noted here rather than hidden.

Sound Match fits every voice at once: same answer, in a fraction of the time

• Analyze runs in parallel now. It used to fit the voice types one after another on a single core; it now fits them at the same time, one worker per physical core. On a fast multi-core machine that is about 4.5× quicker, and the answer does not move by a single bit. The ranking, the scores and every fitted setting are identical to the old serial run, pinned by a byte-for-byte test rather than offered as a promise.
• The speedup is honest about your hardware. The number of workers is your machine's own physical core count, never the doubled hyperthread count, and never more than the number of voices being searched, so the gain depends on how many cores you have. It is bounded by the single slowest voice to fit, which means past a point more cores stop helping and matching one lone voice type gains nothing. Cancel still ends a run cleanly.
• The analysis got leaner, too. Four under-the-hood changes trim the work the same amount whether it runs serial or parallel: reused FFT plans and analysis windows, a lighter loudness measurement, and a quicker pitch read at the start of a run. Every one is byte-preserving: none of them changes a rendered sample.

The whole ranking, in one list

• The "show the types that fit worse" fold is gone. The ranked answer, every searched type, best first, now sits in one list that fills the panel and scrolls only if it runs long, instead of five rows with the rest tucked into a second, separately-scrolling section. Everything was already worked out and sorted; the panel just shows all of it now. A voice that could not reach your sample is a fact worth seeing at a glance, not a click away.

A new front door

• A fresh install opens empty, in the full studio. Instead of somebody else's factory song, a brand-new install opens on an empty song, eight bars with one + Add track prompt, in the advanced view, with every panel on screen. The factory techno starter is still one click away under File ▸ New ▸ From style.
• Three offers, nothing running by itself. The empty launch makes three offers in turn: a look around the window, the guided Build your first song, then a switch to the simplified view. Each is one press, each has a Don't ask again, and not one of them starts on its own.
• The window is bigger, and remembers where it was. A first launch opens maximized; after that RaveFoundry reopens exactly where and how big you left it, checked against the monitors you have now. The declared minimum window is larger, so a fresh song's tracks and the whole transport bar are all on screen at once.
• Every tour steps through its cards one at a time. The guided tours now light one control per card, with Back / Next / Skip and "Step 4 of 12". They put the panel each card is about in place, and put your view and panels back exactly as they were when you leave. Nothing about your song is touched, on any tour.

Panels that get out of your way

• The Mixer is a window of its own, and its faders are full length again. The Mixer opens as its own floating window, every fader back to its full travel with no scrolling to reach one, and opening it now costs the step grid nothing at all. It remembers where you put it and how big you made it, so a second monitor stays a second monitor across sessions.
• Your macros come first. In the track inspector, the four macros you defined now sit at the top of the column, on screen the moment you select a track.
• Find one dial among a hundred and forty. A CHANGED filter hides every knob still at its factory value, and a search box (Ctrl+F) matches a dial's full name as well as the short label under it, so "resonance" finds Res Bass and "filter mode" finds Flt Mode. Float the inspector wide and its sections lay out in columns instead of one long stripe.
• The waveform strip says what it is a picture of. The strip along the bottom now carries a SONG / TRACK chip: SONG is the whole mix as before, and TRACK draws the selected track alone over the same bars, its own voice, rack, EQ and automation, the way a stem export hears it. The status line names which it is showing.

The app says what it is doing

• Stepped dials name their settings. Twenty-one stepped dials read words instead of numbers now, on hover, in the lock readout and to a screen reader. The one that matters most is the bass filter mode, which reads Low · High · Band · Notch · Comb · Ladder instead of a bare 5, with the diode-ladder acid filter named as the sixth position it is.
• Every dial in the effects rack explains itself: all 127, across 21 devices. A rack control used to offer only its label, its range and its default; each one now carries a plain sentence about what it does, written from the engine's own code rather than the manual.
• Every icon-only button tells a screen reader what it does. Twenty-seven glyph buttons used to announce themselves as nothing: move track up, audition a track, the ✕ on a clip, a macro, a lane, a route. Each now speaks a real name, and the ones that repeat per row say which row. Nothing moved and nothing looks different.
• Tooltips that describe the control you are pointing at. A handful of tooltips that named a different feature, or offered a setting the engine ignores, were corrected to say what actually happens under your cursor.
• Text you can read, and controls you can see without pointing at them. The app's small-text tier went up to Windows' own published floor, and combo boxes, fields and pickers now carry a resting outline that clears the contrast standard on every background, where before they had no visible edge until the pointer lit them up.

Sound Match, in colour

• The ranked answer wears colour. Every ranked row now carries a coloured band beside its word and a score bar beside its percentage: green for Excellent, chartreuse for Strong, yellow for Fair, orange for Weak, and a receding grey for a type that could not reach your sample. The colour is a second reading of the band, never the only one: the word stays, the figure stays.
• The bands were re-anchored, with their own rulers. Drums and held notes are judged on separate scales now, so a fit that had not changed a setting is not quietly re-graded when something unrelated moves.
• Nothing offered that cannot be delivered. Dead buttons are greyed until there is something to act on, the four steps are numbered so the panel reads as a walk, and the promises it only ever made in tooltips (that your file on disk is never touched, and every preview is loudness-matched) are printed on the panel itself.
• It still uses your accent. Sound Match keeps the accent colour you chose in the Theme menu; the one fixed hue added is the Strong band's chartreuse, which has to stay put to keep meaning what it means.

Scrolling and dragging that go where you push them

• A selection drag follows the sequencer to its edge. Hold a drag-select at the edge of the grid and the song scrolls under it, in either direction, with the selection extending onto the bars that arrive, so a selection is no longer limited to what happened to be on screen when you pressed.
• A sideways trackpad drag finally scrolls. A two-finger horizontal drag now moves whatever you are pointing at, the sequencer and its lanes together, or the mixer's strips, where before it did nothing anywhere in the app. Your Windows scrolling-direction setting is respected, and a slow drag scrolls slowly.
• No list jumps a whole row per notch. Every list scrolls by the pixel now, so a tall-rowed list, like the MIDI import mapping or the preset library, stops throwing itself past you a full row at a time.

The one sound that changed: the acid accent now makes a sound

• Ticking ACCENT is finally audible. v0.24 built the acid accent five ways over, and every one of them wrote the flag correctly, but the depth dial they all multiply through, COLOR ▸ Accent, shipped at 0, and at 0 an accented step is bit-identical to an unaccented one. The default is now 0.6, chosen by measurement so that ticking ACCENT lifts the note by more than a decibel and opens the filter about half an octave.
• Nothing else moves. An unaccented note is bit-identical at any depth, so an ordinary bass line is untouched, and a fully-tied rolling bass takes no accent at all, because a slid note never re-strikes to arm one. To play an old line exactly as before, set its COLOR ▸ Accent back to 0.

Fixes

• The step inspector's LENGTH field shows its value. A note's span was being set and stored but drawn blank; it reads normally now, and you can still type a value straight in.
• Applying pattern text says what it did. Text that marks no accents comes back with every melodic note accented, which is the parser's rule, unchanged. The status line now names which rule ran and how many notes it touched.
• The step inspector and lane strip talk to screen readers. Both are fully addressable now, and every lane reports what it holds ("2 of 4 steps on") rather than nothing.
• A launch that fails no longer lingers invisibly. A startup that produces no window now closes and reports it, instead of leaving a running process with no interface and a success status.

The acid bass

• A diode-ladder filter. The bass gains the filter the acid sound is actually made of: a four-stage diode ladder with per-stage saturation driven by Grit, calibrated against the classic hardware, running on its own oversampled core. It's a sixth filter mode on the bass, sitting alongside the five it already had.
• Accent is a real step property. Every bass step now carries its own accent, set by a stroke on the cell, a checkbox, a lane or a right-click, instead of being guessed from velocity. A quiet accented note (the thing an acid line does constantly) and a loud one that deliberately doesn't accent are finally one click each, and accent survives a MIDI or copy-as-text round trip. A new Articulation variation flips accents and slides while touching nothing else.
• Six new dials, a sub, and acid macros. Key Follow, Wave Morph, Res Curve, Res Bass, Phase Mode and Env Reset join the bass, all dormant at their defaults, along with the sub-oscillator Lead and Pad already carried, and three documented acid macro recipes (Acid, Bite, Liquid) you dial in yourself.
• A new "Acid Circuit" preset. A tighter, drier squelch that shows the new acid dials off. It is the first sound added to the factory bank in a while, bringing it to 104. Acid Squelch and Acid Dive are untouched.

Sound Match, sharper and honest

• It recognises the ladder. Sound Match now probes a voice's filter shape and its cutoff together, so a bass made with the new acid filter comes back as that filter rather than an ordinary low-pass, and it searches the whole acid lane: grit, slope, accent, morph, sub and more.
• It fits whole phrases. Give it a sample with more than one note and Sound Match renders candidate sequences that mirror your timing and accents, hearing dials a single held note can't expose, then hands you a render of the winning patch playing the phrase, not just one note of it.
• And it admits what it can't recover. Where a dial genuinely can't be read back from a recording, the report says so, with the measurement that settles it. A recovered filter is labelled a best-fitting shape, not an identification, and the report tells you, dial by dial, what it could and couldn't measure.

More to shape and sequence

• The Saturator. A new rack colour device with three transfer curves (tanh, arctan, cubic), dB-calibrated drive, a parallel mix and an asymmetry control for even-harmonic warmth, bringing the effects rack to twenty-one native devices. Dormant at its default.
• A song-wide seed. One number re-keys every probability roll, note-pool pick and humanize wobble, with a Reroll button for instant variations. A set seed makes a variation export-stable, so the same seed always renders the same take.
• The PRE fill condition. Steps marked PRE play the anticipation before a fill. Press and hold Fill partway through a pass and they carry the rest of it, then hand over to the fill on the next pass. It joins the other trig conditions.

Fixes

• The Noise FX "static" is gone. Last release's stereo-balance fix laid a faint hiss under a resonant riser; the voice now plays exactly the sound it always did, centred and clean.
• Parameter locks land on their own step. Gate, Swing and Strum locks now take effect on exactly the step they're written on instead of nudging the following note, and a cell preview now sounds that cell's own locks, rack locks included.
• Accent survives every export. The new accent flag now carries through offline exports, freezes and a MIDI export-and-reimport intact.

Sound Match: experimental

• "What voice makes this sound?" Open Tools ▸ Sound Match, drop in a short sample (up to 20 seconds: a kick, a stab, a pad, one note of anything, trimmed first in the built-in editor if you like) and press Analyze. RaveFoundry fits every one of its own synth voices to your sample and ranks them best first, each with the exact settings it landed on and a plain-English list of what that voice could not reach. Add the closest as a new track, or save it as your preset.
• Deterministic analysis-by-synthesis. Each voice is fitted by thousands of test renders under one search engine, so the same sample always gives the same ranking, scores and settings, on any machine. It matches not just a voice's core dials but its per-track EQ and saturation, the patch's tone, and runs in the background with an honest estimate of how long it will take and a Cancel you can hit any time. No sample library and no machine learning: every answer is RaveFoundry's own synthesis.
• It hears motion, not just a snapshot. The match follows the pitch and the brightness through the note, so a tight dive and a held note at the same pitch no longer score alike and a rising squelch is heard as rising. Multi-note samples get onset markers, the cleanest note is fitted first, and you click another to refit. Every fit previews as a one-hit and as a voice-aware pattern (a rhythm for drums, a phrase for melodic voices) at your tempo, loudness-matched to your sample.
• Two honest read-outs that never sway the ranking. A transposability check quietly re-renders each top fit an octave up and down and reports whether it still tracks pitch exactly and holds its character, because a preset is an instrument, not one note. A "closest existing built-ins" line names the factory presets your sample already resembles.

Deeper sound design

• A second envelope on more voices. Bass, Lead, Pad and FM Bass gain a second modulation envelope for bipolar pitch and cutoff gestures on evolving, sustained sounds, joining the voices that already had one.
• More to shape. A filter attack for a rising acid squelch, longer decay and release tails, its own band for the noise floor and a decorrelated noise width, upper partials that can die faster than the fundamental, an unhardwired ring ratio, and longer percussion decay. At Drive 0 the bass is now a clean sine. Every new control is presetable and automatable like the rest.
• Clap follows the note. A new Key Track dial lets the clap voice track the pitch you play. It sits at its old sound by default and only responds when you turn it up, so nothing you've made changes.

Wider stereo

• Stereo width on eight more voices. The FM family, the wavetable, the cymbals and the tom all gain a true stereo-width control, and the crash shimmer opens up, so more of the kit and the synths can spread across the field, or fold cleanly to mono for the club.

Improvements

• Noise FX sits centred. A Noise FX riser or impact could come out louder in one ear than the other at high resonance; both channels now carry equal energy and differ only in phase, so a sweep lands square in the middle.

The rack grows up

• Fifteen new rack devices. The effects rack fills out to twenty native devices: a compressor, a lookahead limiter, soft clip, a state-variable filter, chorus/flanger, a transient shaper, a multitap delay and a utility, then plate and room reverbs, a phaser, pitch and frequency shifters, a spectral freeze and a granular texture. Each is transparent at its defaults, so placing one adds nothing to the sound until you dial it in.
• Parallel branches and group buses. A rack can split into parallel branches and merge back, phase-honest. And tracks now group onto buses with their own gain, mute, solo and Effects Rack, nesting up to eight deep, so a whole submix goes through one chain, with a per-group stem that still sums back to the mix.
• See what's moving every control. Any control under modulation shows a live indicator with its base value, the range the modulation sweeps, and a real-time marker. A contributor inspector lists exactly what's driving a parameter, in precedence order, with jump-to-source.

Play, map and vary

• MIDI Learn. Map a knob, fader or pad on your controller to almost anything (a rack parameter, a track macro, mute/solo, the transport) in one gesture: click a target, move the control, done. Maps follow the controller between projects or save with the song. RaveFoundry reads your controller; it doesn't drive it back.
• Pattern Variation. Ask for a handful of alternatives under plain constraints, like "keep the kick, 15% busier", "percussion only" or "preserve pitches, alter rhythm". Audition them A/B, and keep the one you like. Every candidate is graded and shows exactly what changed before you listen; it's deterministic, varies your own material, and learns nothing. One undo reverses it.

Master the low end

• The Rumble Generator. A kick-triggered rack device that lays a tuned, distorted, decaying sub tail under your kick, phase-safe for a club's mono subwoofer, with Sub Tail and Room Rumble presets. It's a synthesized resonant tail, not a sample, and at rest it doesn't touch a thing.
• The Low-End Lab. A measurement bench for the bottom octaves. Are the kick and bass fighting, and in which band? Are they phase-aligned? Where is the sub energy? Read it on screen or from the command line, offline, touching the audio engine not at all.
• The Whole-Song Waveform. See the entire song as one strip: an exact picture of the master, aligned to the bar ruler. Click anywhere to seek. It refreshes edited bars in the background and never interrupts playback.

Fixes you can feel

• Live edits land immediately. Undoing a rack knob, or drawing a new ducking shape on the Gain Shaper, now reaches the playing device the moment you make the edit. No save-and-reload to hear it.
• Rack modulation is fully deterministic. A rack with a live LFO, random source or envelope follower now renders identically at every buffer size, so live playback, an export and a seek always agree.
• Cleaner imports and steadier MIDI. An imported preset takes a fresh identity so your bank can't hold two with the same one, and a device fault during a panic now surfaces instead of leaving the keys quietly dead.

A device rack on every track

• The Effects Rack. Every track now carries a device chain. Drop in a three-band EQ, Color, Stereo Width and the new Gain Shaper, reorder them, and every knob on every device is a first-class parameter you can step-lock, automate, drive from macros and modulate.
• A modulation matrix. Route LFO curves, random sources and an envelope follower onto any rack parameter, so a filter can breathe on a curve you drew, or a device can ride the track's own level. It stays deterministic, so an export is bit-for-bit the render you heard.
• The Gain Shaper. Sidechain-style pumping with nothing to patch: pull a track down on a trigger (another track's notes, even a muted ghost track, a beat division, or an audio threshold) and let a drawn curve bring it back. Classic Pump is one click, beat-synced shapes hold their length when the tempo moves, and it adds zero latency.

Keep what you played

• Play it, then keep it. Live input let you play your tracks from a keyboard; now RaveFoundry remembers. While you monitor, the last stretch of your playing is held in memory, with no arming and no red button, so when a run comes out right you can keep it.
• Tools ▸ MIDI Takes. A take arrives split into one candidate per MIDI channel. Hear it back two ways, quantized through the destination track's own voice or exactly as you played it, re-quantize it to a new tempo, and the bin keeps your last sixteen.
• The drop is the only thing that writes. Drop a take onto the selected track, as new tracks, or onto your routed tracks. Each is a single undo, and the result is step-for-step identical to importing the same notes as a MIDI file. Monitoring on its own still writes nothing at all.

Fixes that make it honest

• Stems and frozen tracks now sound like the track. A stem or a frozen clip used to drop the track's own parameter locks, automation, curve-LFOs, macros, cross-track routes and loaded wavetable, so a soloed part could come out wrong. One copy routine now carries everything. Your full mixes are unchanged; re-export any old stems that used those features.
• MIDI input heals itself. When another program wedges Windows' legacy MIDI, RaveFoundry now switches to the modern Windows MIDI Services backend on its own instead of needing a restart, and every dialog and report says which transport is in use.
• Cleaner live-MIDI edges. A device that vanishes is noticed within two seconds, PANIC and a keyboard System Reset make a clean boundary, and a re-arm always comes back healthy. All from a seventeen-item review of the live paths.
• Flat waveforms fixed. The Audio Lab and Sound Designer no longer collapse a whole file onto its first fraction of a second in their waveform views.

Live input

Live MIDI input for class-compliant controllers. Hardware-in-the-loop validation with the Novation Launchpad Pro MK3 is in progress.

• Play your tracks from a MIDI keyboard. Plug in a class-compliant controller, open Tools ▸ MIDI Input, tick Monitoring, and the selected track plays under your hands, with the transport stopped or running, held chords and all.
• Sixteen channels, sixteen tracks. A MIDI channel is an instrument, so up to sixteen of them can drive sixteen different melodic tracks at once, or a single row can follow whichever track you've selected.
• A PANIC that is always there. One button releases every live note on every track and is never greyed out; your sustain pedal, All Sound Off and All Notes Off are all honoured.
• Honest latency. Two extra-low buffer rungs, 10 ms and 15 ms, are offered for playing live, with a late-callback counter that tells you the moment your machine can't hold one. The shipped default is still Safe (100 ms).
• Monitoring, not recording. Live input plays your tracks and never writes to your song. Your live notes and the sequencer's stay completely separate.

A better guided tour

• The callouts got readable. Bigger type, a placement that keeps each card off the control it is naming, and a clean cut-out so that control stays clickable, with a tapered leader in place of the old arrowhead.
• A new "Build your first song" tour. Nine steps, one at a time, in the manual's own order: add a track, set the tempo, lock the key, pick the Draw tool, draw the beat, press Space and listen, give a track its sound, grow the loop into a song, then export it. It tells you what to press; it never touches your song.

Your work, protected

• Reordering tracks keeps your locks. Moving tracks around in Pattern mode no longer retargets or drops your step parameter locks.
• A new song can't discard your recovery copy. Opening or starting a song now flushes the current one's recovery to disk first, so the autosave promise holds.
• Autosave never goes backwards. An older periodic autosave can no longer overwrite the newer snapshot taken at shutdown.
• A busy file no longer loses a clip. A clip whose audio is momentarily locked is kept to heal itself next time, instead of being dropped.
• A newer file is refused, not half-read. A song saved by a future version of RaveFoundry is turned away by name rather than migrated destructively.

More fixes & safeguards

• Locks land on the right step. Level, pan and send locks now fire exactly on the step they were written on. Songs with no locks render identically.
• Chords release cleanly. A Chord track now releases the voice that actually played the note, so a chord no longer cuts early or hangs on.
• No ghost sidechain duck. Toggling the detector's high-pass off no longer fires a duck from stale audio when you switch it back on.
• Content packs are checked, not assumed. A pack has to declare what it is and which format version it is written in, so a file that merely looks like a pack is refused, and an import that fails partway now rolls back completely, library and presets and all.

Getting started

• Press Play and it's there. The song starts the instant you press Play, with no wait for sound, on every press after the first. Pause still cuts dead, and resume is still seamless.
• A simpler first view. Beginner view opens with just the essentials: the transport, the grid and its tracks, the core knobs. The full studio is one menu item away, and every panel returns exactly as you left it.
• The app shows you around. An on-screen tour draws an arrow from each tip to the real control it's naming: a short one for the simple view, a full one for everything else. Re-run either any time from Help ▸ Tour.
• A blank canvas. A new song now starts empty, with no tracks to delete first and one obvious "+ Add track" prompt to place your first one.
• An About box that names your build. Help ▸ About now states the product, the exact version and a link to ravefoundry.com, read from the same source the manual is stamped from, so it can't drift.

Making music

• Hear a preset before you commit. A ▶ beside the preset picker plays the highlighted preset through the track's real engine, with no apply and nothing to undo. The inspector says what your track is actually on, marking it when you've tweaked it since. A RECENT list keeps the last ten you applied per track type.
• Drops you can move. Drag a drop along its lane (snapping to the beat, or hold Alt for step precision) or type its position. Every drop mode now has its own lettered, colour-coded chip, and the lane finally says what it is.
• Your own track groups. The fixed folders are gone: you make the groups, name them, fill them and reorder them, dragging a track from one to another, and every change saves with the song and undoes by name. Miss the old layout? One click rebuilds it as ordinary groups.
• Hear what a click would write. Hover an empty cell in either grid and it previews exactly the note a click would place, scale lock and snapping included, never during playback.
• The Sample Kit works. Load, auto-map and edit velocity zones: Add or Replace on an existing kit, delete and reorder zones, and per-kit velocity-to-cutoff and velocity-to-decay.
• Every knob explains itself. Hover any dial for a small card: its name and one plain sentence about what it does to the sound. Every catalogued dial is covered, and screen readers get the same sentence.

Not losing your work

• Closing with unsaved work finally asks. Save, discard or cancel: a real three-way question, with autosave still backing everything behind it.
• An edit history you can jump back through. Every undo step carries a written name; History ▾ lists the recent ones and one click takes you back to the point you pick.
• Open Recent, by name. Your songs show as "Warehouse Dawn", not a filename, newest first, on the menu and the welcome strip. Plus Ctrl+O to open and Ctrl+E to export.
• The Audio Lab, one gesture away. Right-click an audio clip ▸ Open in Audio Lab, with open-and-slice in one gesture, Ctrl+D to duplicate, and slices that can land as timeline chops or kit zones.

Reachable by everyone

• Full keyboard control, and you can see where you are. Every control takes focus, Space presses the focused button, and the focused control is ringed in your accent colour. That ring appears for keyboard focus only, so clicking never fills the screen with outlines.
• It talks to screen readers. Grid cells, track rows, pitch-grid notes, meters, drop chips, the whole Sound Library and the tour's own cards announce themselves with real names and states.
• Make the whole interface bigger. A single zoom scales every window, pop-outs included, from 75% to 150%.
• An honesty pass throughout. Controls that do nothing on an instrument are removed rather than left to be turned, the status line says what actually happened, and the EQ's Analyse admits when it's looking at the whole mix rather than the band you soloed.

New sounds

• Two mallets join the bank, and the count reaches 137 presets. Marimba and Glockenspiel arrive as struck-body Forge presets: the Marimba dark and woody with a tuned ring, the Glockenspiel a bright metallic onset resolving into a long clean shimmer. A MIDI track named "Marimba" or "Glockenspiel" now imports onto the matching voice.

New effects & features

• Solo any EQ band. Every band on every surgical EQ, track, master and pop-out, now has an S. Hear only the material a band sits on, so you find the boxy 400 Hz or the harsh 3 kHz directly instead of inferring it from how the whole track moves.
• Mid/side, per band. Each band can act on the whole stereo image, the centre only, or the sides only. Take a box out of the middle of a wide pad without narrowing it, or lift air onto the sides while the lead down the centre stays put. The curve splits into two traces to show which is which.
• The master EQ can show you the mix. Press play, then Analyse, and the master EQ draws the live output you are actually hearing, with a second curve behind it showing exactly what your bands are doing to it. Nothing is re-rendered.
• Broadcast-standard metering. A live EBU meter in the Master Lab, with momentary and short-term loudness and a true-peak read that warns above 0 dBTP, plus a loudness-range figure on every export, beside the integrated LUFS.
• A second envelope for Wavetable. Env 2 now reaches the wavetable's own morph position, the sweep a wavetable synth is for, joining the melodic voices that already had it.
• The endless riser, everywhere. The Shepard riser now drives the Transition Baker's new "Endless Build", a Rave-Mode pad and both auto-arrangers, and a CYCLES control sets how many times it climbs inside its window, so the wrap lands where you can hear it.
• Revert a clip to its source. The Audio Lab keeps a clip's original audio when you bake over it; one button in the clip inspector brings it back, even for a fresh recording that was never saved separately.

Workflow

• Convert to Patterns cuts where the music does. Turning a song into patterns now cuts at your drops instead of into anonymous 8-bar tiles, so a long track comes out as its intro, build and breakdown, a map you can read. A song with no drops converts exactly as before, and it is still lossless both ways.
• Find what things sound like. Pick any sample or preset in the Sound Library and Find similar re-orders the rest by how close they actually sound, on-device, deterministic, nothing uploaded. Every filter you had keeps working inside the ranking.
• See a pack before you let it in. Import pack now shows you every sample and preset first, with sizes, lengths, what you already have and what would collide, then lets you keep yours, keep both, or take the pack's names. Nothing is touched until you say so, and Undo puts it all back.
• Your library, where you want it. Point the Sound Library at any folder, an external drive or a shared location, and back again, with nothing moved or deleted. And it now suggests tags: it listens to each sample and proposes one-shot / loop, dark / bright, pitched / noisy hints you can accept or ignore.
• Your mod wheel survives the import. A MIDI import now brings its controller performance in as automation curves, mod wheel to cutoff and expression to level, one lane at a time, each shown before it lands.
• Your Tempo Map leaves with the .mid. MIDI export now writes every tempo change out at its exact bar, so a song's tempo map round-trips out and back unchanged.
• Three step-grid sizes. A CELL strip sets the grid to S, M or L, a dense whole-pattern overview or a large touch-friendly view, remembered across restarts, without changing the grid's resolution.

Improvements

• Chords play as written. A Chord track now voices exactly the notes you authored instead of rebuilding a chord on top of each one, and holds up to six tones. What the pitch grid shows is what you hear.
• Pattern locks play, and survive. A parameter lock inside a pattern now sounds when you audition it, and stays put through saving, leaving Pattern mode and editing, instead of quietly vanishing.
• Undo returns the song exactly. An undo that changed the song's structure used to nudge every kick's starting phase; now the song comes back bit-for-bit, however many times you undo and redo.
• The Fractal Echo lands on the grid. After a tempo change the echo tree now settles exactly on the grid instead of a hair short, and its capacity warning appears while you drag the tempo, not only on load.

New sounds

• Nine new sounds, and the first tom family. The Tom voice grows to Low / Mid / High Tom, three tuned sizes so a fill reads as one kit, alongside a Square Lead, Analog Brass Stab, Synth Vox Pad, Digital Vibraphone, Cathedral Drawbars and Digital Harpsichord. The factory bank is now 135 presets.

Writing

• The pitch grid. Write melodies where the melody is. A piano-roll-style pitch view for the melodic voices, sharing the same steps, locks and automation as the drum grid you already know, so a bassline or a lead is drawn, not typed into a note pool.
• Step locks on every kind of track. Parameter locks now work on audio and vocal tracks too, not just the synths. Any step, any parameter, any track type.
• Copy a pattern onto any track. Copied steps now paste across track types instead of refusing, and if you know what you are doing, the overwrite prompt can be told to stop asking.
• Change tempo without orphaning your clips. Move the song's BPM and your audio clips keep their musical position instead of drifting off the bar they belonged to.
• Tempo Map: change tempo mid-song. Set a starting BPM, then add stepped changes at any later bar, for the half-time breakdown or the second-half lift. Everything that follows the beat moves with it, and the song still renders the same every time.
• Timing Lanes & Track Clock. Each track can read its own grid rate, from 1/8× up to 2×, for half-time hats, double-time arps and polymetric layers over one shared tempo.
• Chords now hold six notes. Chord steps sound up to six tones, with six-note pattern text and a MIDI importer that keeps full voicings whole instead of shaving them to four.

Mixing

• A surgical EQ on every track. Up to eight bands placed anywhere from 20 Hz to 20 kHz, as bells, shelves, pass and notch, drawn on a curve, with Analyse overlaying the track's own spectrum.
• Automate the surgical EQ. The eight-band EQ can now be automated as a whole block. Sweep the correction through a breakdown, or open it up at the drop.
• An EQ for the whole mix. The same surgical EQ now sits on the master bus, placed before the master dynamics so everything downstream hears the corrected mix.
• A better EQ panel. Labelled frequency and dB axes, typed number entry for every band, a live read-out, and a pop-out window when the inspector is too small.

Exporting & sharing

• MP3 exports of any length. No temp files and no size ceiling: a two-hour set exports the same way a two-minute loop does.
• Stems render all at once. A stems export now renders every track in parallel instead of one after another.
• Quality checks that read the whole file, and your stems. The export QC no longer samples the opening; it reads the finished file end to end, covers each stem as well, and says plainly what it found.
• Send someone your sounds. Package samples and track presets into a single content pack and hand it over. Every pack carries its own licence guard, so nothing arrives without knowing where it came from.
• MIDI in and out, in full. Import a multi-track .mid as an editable arrangement, setting each track's type and preset before committing, or export your song as General MIDI: drums on channel 10, chords kept intact, and each melodic voice on its own channel with its instrument.

Improvements

• Loading a preset gives you that preset. A preset now replaces what was on the track instead of stacking on top of the last one, and it leaves all your sends alone, not just some of them.
• MIDI import tells you what it changed. Importing a file reports exactly what the 16th-note grid did to your notes, instead of quietly moving them.
• Octave-robust Hard Tune. The vocal tuner now holds the intended note on wide, formant-heavy takes instead of jumping an octave.
• Reachable by assistive technology. Every dial and step cell now reports its name and value to screen readers, and can be read and set through assistive tech, not just clicked with a mouse.

New instruments & sounds

• Forge Body: a struck-body instrument. A physically-modeled voice: strike a free bar, steel plate, bell, membrane drum, woodblock or steelpan and hear real geometry ring and decay. Struck metal, wood and skin, modeled rather than sampled.
• Sample Kit: a true multisample instrument. A new track type: lay samples across zones, each with its own key range, velocity layer, tuning and choke group. Velocity layers crossfade, zones repitch per note, and open/closed pairs choke like a real kit. Build one in a click with chromatic auto-map, or send Audio Lab slices straight to the zones.
• Pluck, deeper. Four new character dials (a wooden body resonance, real pick noise, a rounder excitation blend, and independent brightness and decay) take the string voice from harp to nylon. Existing presets sound exactly as before until you reach for them.
• Cinematic sounds. New factory presets for scoring the drop: Cinematic Boom, Threat Drone, Cinematic Low Drone, Tonal Descent, the Air Whoosh / Reverse Whoosh pass-bys, and Cinematic BRAAAM, a big blooming brass impact. Noise hits can now play a whole riser or whoosh from a single trigger.

New effects & features

• Weave: generative sequencing. On any melodic track, each note sprouts a branching tree of new notes, transposing and fading with depth, all played by the track's own voice. An arp that grows rather than cycles, and it weaves the same way on every loop and every export.
• Fractal Echo. A new send alongside delay and reverb: one echo spawns generations of self-similar taps, each darker, wider and quieter than the last, up to 64 taps of recursive echo from a single dial, shaped with ratio, tone, spread and decay.
• Future Shadow. Pre-echo ghosts for vocal and audio tracks: the track plays what's about to happen, so a hit swells in before it lands. The drop hears you coming. Modes: Echo, Reverse, and Vocal Ghost, a whispered lead-in.
• Hard-tune vocals. A new stage in the vocal chain: pitch correction that snaps to your key and scale, from a transparent nudge to full instant-robotic, with an expression-preserve dial that keeps the voice's vibrato and scoops.
• Envelope character + a second envelope. Per-stage curve dials for attack, decay and release, soft to snappy, on every envelope voice, plus a second envelope on FM Synth, FM Keys, Chord and Granular you can point at pitch or timbre for evolving patches.
• Shepard riser drop. A new drop mode: an endless-rising riser that seems to climb for the whole drop without ever running out of range. It is the classic Shepard illusion no ordinary riser can fake.

Workflow

• An export that preserves your master. A plain export no longer quietly raises your level. A new Ceiling picker offers Preserve master (the new default, where the file is exactly what you rendered), Peak-normalize, and Loudness target, and the status line reports the exported file's true peak and loudness.
• One-click auto-slice. One button in the Audio Lab finds the hits, takes the loudest, and lays them onto sampler zones: the whole slice-to-kit chain in a single press.
• Parameter locks on every parameter. Step locks now take effect on every control in the app: pluck tone, the hard-tune dials, envelope curves, the sends and more.

Improvements

• Honest names across the bank. The pluck voice is a bright string / harp machine, so its presets say so: Nylon → Nylon Pluck, Steel Harp → Bright Harp, Palm Mute → Muted Stab, Power Stab → Grit Stab, Industrial Chug → Dark Grind, Driven Sustain → Driven String. Every renamed preset is the exact same sound. Koto is retired, and the vocal "Room" reverb is now Vocal Plate.
• Cleaner tone and honest metering. The mechanical organ key-click is gone, the string pick-click is softened, and Master Lab's loudness and true-peak meters are now conforming, externally-validated instruments with full-song spectrum and stereo analysis.
• Refined vocal effects. Beat Repeat splits into Chance, Depth and Division; Shimmer is a true reverb send that no longer thins the dry vocal; and the vocoder, vowel filter and transient shaper are smoother.
• Sound polish. Dozens of factory presets were re-voiced against hardware references, among them a controlled sub snare, an evolving FM swell pad and cleaner acid and hoover. New songs start with better-calibrated levels.

New effects & features

• A new master limiter. A clean, purpose-built master limiter with look-ahead and a true-peak ceiling, so loud passages stay controlled without the old harsh crunch.

Improvements

• A tighter low end. Stacked kicks no longer pile up perfectly in phase, so dense drum patterns stop overloading, which was the root of the old low-end distortion, and the most over-driven demos were dialled back for a cleaner bottom.

New instruments & sounds

• Strobe Cathedral. A 13th bundled demo and the set's finale: neo-rave hard techno. File ▸ Open Demo lists all 13, and a demo always opens as an unsaved copy.
• Hoover Classic + reference percussion. The archetypal rave stab from a classic analog-synth recipe, plus a new rimshot and a retuned cowbell voiced from hardware references. The bank grows to 105, with a refreshed set of style starters.

New effects & features

• Engine polish. Reverb gains pre-delay and tail modulation, the multiband gets per-band depth, and Noise FX can self-oscillate.
• Spectral vocoder + transition baker. Preview the vocoder morph before you commit and drive it straight from a WAV/MP3/OGG clip; the transition baker now lets you pick the target bar and preview before baking.

Workflow

• Transport that behaves. Play/pause holds the playhead and resumes where it stopped; Start Over returns the playhead and view to bar 1.
• Panels. A View ▸ Panels menu with persisted show/hide and consistent float/dock, a step legend under the sequencer, and a sticky manual table of contents.

New instruments & sounds

• Hoover, rebuilt. The Lead gains a PWM-morphing saw and a per-note attack pitch-swoop: the classic hoover / rave stab.
• Acid bass filter. In-loop resonance grit for acid snarl and self-oscillation, off by default, so existing basses are unchanged.

Workflow

• A bundled, offline manual. Twelve chapters from getting started to troubleshooting, a guided first beat and first song, and a full reference, all under Help ▸ User Manual, with no network needed. Every screenshot is generated from the live app, so it always matches.
• F1 cheat-sheet + menu polish. A one-page keyboard-and-gesture sheet from anywhere, plus checkmarked panels, a clean icon set, and confirm-on-delete for snapshots and takes.