SoundCurrent DAW

A Linux-first professional digital audio workstation and recording suite, built with C++20, CMake and Qt 6, licensed GPL-3.0-only. The processing engine is independent of Qt and shared by live audio and offline rendering.

Early development preview — not a production-ready DAW. The target is the professional composition, recording, editing, mixing, sound design and live performance workflows in SC-DAW-BASELINE-2026-10-05. No matrix family is currently qualified as equivalent. Functional parity, processing quality, bundled content and native-project compatibility are tracked separately.

Windows functionality, all-Europe localization and imports from other suites are required parts of the goal. Native Windows audio/desktop/installers and reviewed translations remain unfinished. See the active goal and the acceptance matrix.

Project-import development now includes approved media folder/file handles and pinned WAVE metadata/content checking. The desktop inspector now has an explicit media checklist with folder/replacement selection, format/sample-peak results and cancellable child checks. Original source bytes and floating headroom are preserved. Broader audio formats and destination project conversion remain required; installed previews need a refresh and separate qualification.

Development also includes exact positioned clip playback/export for owned assets at different physical sample rates. The desktop shows exact fractional source positions and offers project-frame crop/split with Undo/Redo. An isolated pitch/time-stretch worker now renders owned source spans to floating RF64 with independent pitch and duration, resource limits and cancellation. Editable raw-anchor state and checked artifact adoption retain raw media and derived assets through split/crop, re-render and Undo/Redo. Full processing-quality and current native qualification remain open.

The clip playback-speed control retimes duration and fades with exact rational state and Undo/Redo; pitch follows speed. The clip pitch/stretch dialog now provides background rendering and explicit verified, undoable adoption. Native Windows build/process qualification and a local installer/source pair now include the required helper. Installed Windows setup and startup pass payload, shortcut, Unicode project-open and normal-exit checks. The installed Windows editing/export workflow also passes render/review/Apply, Undo/Redo/save, new-process reopen, identical repeat WAV exports and removal/reinstallation preservation on an owned synthetic project.

A separately identified unity render now preserves prepared source samples at duration 1/1 and pitch zero, including one-frame clips. Existing rendered assets retain their original algorithm identity. Current-source Windows and installed acceptance for this increment are independently scoped. An explicit source-context workflow now lets short clips use chosen neighboring audio, with nominal visible duration, restored controls, Undo and shared playback/export acceptance. Acoustic transient alignment and current-source installed qualification remain open.

Retained quality diagnostics now include 63 full synthetic renders and 37 crop comparisons. Unity copies remain exact; non-unity output depends on processing context. Local attack timing and grouped microphone phase remain explicit gates before warp-marker delivery.

An experimental protected-marker editor now prepares verified derivatives with audition before Apply, Undo/Redo, save/reopen and shared playback/export. Audition uses an ephemeral project model and temporary output choices. Integer/spacing/pitch/context limits are explicit; sustained grouped-phase and broader quality, native and installed gates remain open. This is a source development workflow, not a newly qualified installer.

Full pitch/stretch quality, foreign-property adoption and installed/native audio qualification remain open. Installed Linux evidence covers its recorded earlier source. Short-span/transient quality, tempo/warp editing and sustained native recording remain separate gates.

The Windows desktop workflow now records through the native desktop, plays with live EQ/Undo, saves/reopens and exports WAV. Independent raw/output/export sample checks pass on an owned loopback route with a silent lead-in. A local Windows 11 x64 installer/source pair now passes scoped clean-install, shortcut, normal removal/reinstall and installed native workflow acceptance. Bounded non-silent startup and normal endings now have independent native evidence. Physical/sustained recording and monitoring/duplex remain open preview gates. The refreshed Windows installer passes normal installation/UI checks, but all three installed native attempts detected capture discontinuities. It is an installation/UI preview with experimental audio; the saved full/partial raw takes and failures remain retained. Abrupt production EQ Stop also remains unresolved.

The bounded Windows capture trace retains one complete and two refused recordings, with exact raw samples and SDK packet gaps. It is diagnostic evidence; installed recording reliability remains open.

Easy installation is required on supported Linux and Windows systems: normal packages/installers, no compiler or manual dependency assembly, app-menu shortcuts, and tested upgrades/removal that preserve recordings. The build instructions below are for developers; broader end-user installer qualification remains open. The preview delivery plan starts with the existing Linux recording/EQ/project/WAV workflow. Local binary packages need fresh-machine qualification before being described as supported installers. An Ubuntu 26.04 amd64 DEB candidate now passes scoped runtime dependency installation, normal-user project/export and upgrade/remove/reinstall checks in an owned Ubuntu Base environment. See the candidate guide. The installed app also passes a scoped owned-source ten-second recording, live EQ/Undo, save/reopen and independently checked WAV export workflow. Recording startup/clock diagnostics, full desktop/menu and physical/sustained recording qualification remain open. Single-track code now retains portable recording error details for reopening. The latest local Ubuntu candidate passes an installed fault/Save/Quit/reopen/review workflow. Sustained physical recording and native Windows qualification remain open. The same binary now passes normal 11.264-second recording, live EQ/Undo, save/reopen and verified WAV export.

This is the DAW repository. soundcurrent-eq is the free equalizer; soundcurrent-studio is the separate premium equalizer. Borrowed components have pinned provenance and are adapted here without changing those repositories.

What works today

• Versioned portable project state, stable IDs, strict validation, worker-side save/reopen and typed/grouped Undo/Redo.
• Prepared in-process EQ, smoothing, bounded live parameter events and floating-point headroom.
• WAV/RF64 recording, read-ahead playback, verified take admission and checkpoint recovery, and offline WAV export through the shared processing engine.
• An opt-in Qt desktop with tracks/timeline, explicit Linux PipeWire routes, live EQ, colored meters, master channel matrix and simultaneous armed recording with project playback.
• Saved punch locators and grouped takes, with short 32-track owned-native desktop evidence for raw samples, live EQ output, Undo/Redo and Save/reopen.
• Equipment profile browsing, imports and editable custom copies from a pinned 1,092-speaker catalog. Monitoring/printing correction and portable profile pins still need implementation.

These workflows have scoped test receipts. Sustained native recording failures remain open. Synthetic or short owned-route success does not establish physical latency, a supported Windows application, or full DAW parity. The changelog summarizes the development snapshot. The localization refresh adds reviewed equipment/settings vocabulary and signed RTL input; 33 language catalogs remain partial drafts with English fallback, awaiting native-speaker review.

The product targets lower-budget recording studios, including studios with substantial hardware. Track scalability requires no fixed product/license ceiling on total project tracks, with separately measured real-time and hardware-input capacity. The first scaling implementation removes the fixed total-track ceiling from the model, parser, mixer and desktop Add Track control. Synthetic 4,096-track core and Linux Qt workflows have scoped evidence. A shared media pool/cache now serves 1,024 file-backed tracks with one handle in a bounded exact-output test. Undo resource settings let the desktop raise retained command, payload and operation-workspace budgets. Existing history survives rejected reductions; usage and oldest-command retirement are visible. These declared checks are one input to future combined memory admission. The retained-session ledger also keeps older immutable snapshots charged until their last reader releases them. Saves and barriers share current state; snapshot-budget refusal preserves active edits and leaves Cancel available. Shared controller admission now also charges canonical state, history and declared edit work against one parent. Edit → Project resources… exposes persistent adjustable parent/snapshot policies and live usage. GUI payload admission now leases selected-track copies, lists and timeline indices from that parent. A refused display retains its old view with resource settings and explicit retry. Prepared execution admission extends that parent to DSP/playback pools, reader buffers, shared caches and WAV output buffers; old/new graphs retain credit through control-side retirement. Coordinated execution policy makes new desktop playback/recording/export preparation follow the trusted parent and adds declared capture pools, take banks, bridge bindings and writer workspace. Copied owner state, other IO, exact allocator/RSS and measured sustained capacity remain required.

The viewport timeline and model-backed lists now have bounded Linux Qt large-project regression workflows. Combined memory, recording, meters/waveforms, freeze/bounce and sustained Linux/Windows workloads remain staged.

Build on Linux

Core dependencies: a C++20 compiler, CMake >=3.20, OpenSSL 3 Crypto development files, and libsndfile 1.2.2. JSON and the libsndfile API header are vendored with third-party notices. Desktop builds also need Qt >=6.4 Core/Gui/Widgets; fixtures need Qt Test. Ninja is used in these examples.

Native Linux audio requires pkg-config and PipeWire development headers >=1.6.2 with an existing user daemon. Build products are kept in ignored .cache/.

cmake -S . -B .cache/build-desktop -G Ninja \
  -DCMAKE_BUILD_TYPE=Debug -DSC_BUILD_DESKTOP=ON
cmake --build .cache/build-desktop --parallel 2
ctest --test-dir .cache/build-desktop --output-on-failure --parallel 1
.cache/build-desktop/soundcurrent-daw

For development machines without that PipeWire version, add -DSC_BUILD_PIPEWIRE=OFF. This permits editor, synthetic and offline workflows; native playback and recording are unavailable in that build. No PipeWire fork or replacement daemon is bundled. With SC_BUILD_DESKTOP=OFF (the default), the core build requires no Qt. SC_BUILD_MEDIA=OFF selects state/engine/transport only.

CTest does not connect physical audio devices. Native integration fixtures are opt-in and must run serially after builds and tests finish.

Desktop recording workflow

Create/open a project, select or create a track, prepare recording and explicitly choose its input/output routes. Arm the selected track and use Record/Stop. Successful raw takes are verified before attachment. To overdub multiple tracks, enable Record armed tracks with project playback, select arms, prepare and choose all packed inputs/master outputs explicitly. Punch settings are captured at preparation and cannot change while the prepared session is active.

Live scalar EQ edits and Undo use applied-frame receipts. File → Recover recording… previews/copies an owned checkpoint; File → Export WAV… renders an immutable project snapshot with range, cancellation and overwrite confirmation. Closing waits for recording/export finalization before Save/Discard/Cancel.

See desktop recording, multi-arm workflow, punch controls, and export for bounds and remaining work.

Developer tools

.cache/build-desktop/sc-project-tool new .cache/example-project
.cache/build-desktop/sc-project-tool inspect .cache/example-project
.cache/build-desktop/sc-record-tool synthetic .cache/new-synthetic-project
.cache/build-desktop/sc-export-tool render-mix PROJECT_DIRECTORY MIX.wav

synthetic generates input and uses the real disk worker; it refuses an existing project directory. Export refuses existing targets without explicit hash-based consent. These commands are development tools, not a complete user installation. See recording/recovery and offline export.

Windows source and development status

Clone the same repository on Windows:

git clone https://github.com/rhamenator/soundcurrent-daw.git
Set-Location soundcurrent-daw

Native MSVC/Qt Windows desktop and owned WASAPI workflows now have separate acceptance evidence. A per-user Windows 11 x64 installer and matching GPL/dependency sources are being prepared locally; clean-installed acceptance is a separate gate. On a Linux host with MinGW:

cmake -S . -B .cache/build-windows-core -G Ninja \
  -DCMAKE_TOOLCHAIN_FILE=cmake/toolchains/windows-mingw.cmake \
  -DSC_BUILD_MEDIA=OFF -DCMAKE_BUILD_TYPE=Release
cmake --build .cache/build-windows-core --parallel 2

scripts/build_windows_media.py additionally builds hash-pinned libsndfile 1.2.2 for the headless disk layer; it does not qualify Windows runtime or deploy all runtime dependencies. The planned native adapter uses user-mode WASAPI. Source copies and unsigned local development do not require buying a signing certificate. Trusted distribution and installers remain separate gates; see Windows/localization and signing budget.

Plan, evidence and next task

The native and installed protected Windows preview now has retained marker-render, Undo/Redo, Unicode reopen and float export evidence. The refreshed unsigned installer and corresponding source are prepared locally; physical audio, pristine Windows installation and full-suite qualification remain open.

Retained render review adds explicit background scan and review after reopening a project, with source/media verification, stale-target guards and transactional Apply/Undo. Installed restart qualification and scalable lifetime job management remain open.

• Scope and frozen baseline
• Source-linked parity matrix · machine-readable requirements
• Architecture and real-time threading contracts
• Dependencies and licenses · decision records
• Milestones and measurable exit criteria · first-slice gates
• Other-suite project import · equipment profiles/editor
• Continuing equalizer reuse audit
• European language coverage audit
• Native desktop punch evidence · punch interruption/recovery
• Main branch protection and PR workflow
• Track scalability without a fixed product ceiling
• Complete documentation index · dated test results

Saved per-track input-latency controls now connect canonical project state to accepted recording preparation and take alignment. Saved recording-only Auto monitoring selects live input inside desired punch boundaries while preserving continuous EQ history. Prepared deferred capture starts now let audio publish an exact one-shot origin before disk startup, with immutable worker binding. The manual punch engine owner now preserves continuous playback/EQ across replenishable takes, reliable commands and delayed postroll, with scoped synthetic disk/concurrency acceptance. The manual recording control owner now starts and joins consumers, verifies bounded groups and handles empty/faulted takes. The native manual owner now has finite 32-arm repeated-take and late-service acceptance on owned PipeWire routes. Default exports are repeatable across render seconds. Native manual fault/recovery tests now expose retained channel-alignment and callback-deadline failures alongside finite owned-route successes. Priority Stop/Cancel now terminates audio independently of held disk startup, with scoped native Linux evidence. Public native buffer handoff observations now retain IO/buffer startup state beside independent waveform markers. One new finite Stop passes; original46's delayed channel remains unresolved. A controlled startup experiment now reproduces a one-cycle delay when an allocated buffer is queried before IO readiness; deferring that query prevents the measured mechanism. Production readiness and capacity-aware acquisition retain logical port tracing. The manual recording desktop panel now connects finite repeated takes, explicit routing, EQ updates, preview adoption and priority Stop/Cancel to Qt controls. Owned native desktop evidence now checks repeated groups on3/32 inputs, live EQ, adoption/retry and Stop/Cancel/Close. Valid native silence is handled without exposing stale backing bytes. A separate active clock gap under sanitizers remains unresolved; sustained/physical and Windows runtime qualification remain open. Additional monitor policies, tempo/loop/take lanes and comping remain required. Native engine punch fault recovery now has scoped evidence; desktop fault/discovery, process-kill and empty-preroll workflows remain required. Sustained native recording, independent Windows qualification and all other frozen-reference requirements remain required. The first recording slice still has open independent acceptance gates.

Contributing, security and license

See CONTRIBUTING.md and SECURITY.md. GitHub CI builds/tests Linux synthetic desktop workflows without a native audio daemon and cross-compiles Windows core targets. These checks do not claim native device or Windows runtime qualification. No hosted release binaries are published yet.

GPL-3.0-only permits paid distribution while preserving recipients' license rights and corresponding-source obligations. Borrowed code/data retain notices and provenance; proprietary assets and algorithms are not assumed reusable. No dual-license rights are implied.

GitHub backs up tracked source, history, documentation and committed receipts. Build caches, local recordings and ignored failure executables/media are excluded; back those up separately if needed. Windows copy and backup instructions.

The development foreign-project inspection shows original REAPER property values and preserved source lines, including missing/unsupported state. Saved inspections reopen without the original file; importing their audio into a destination project remains under development.

Editable clip gain and fades (development source)

Select a clip and expand Clip gain and fades to set gain, mute, polarity and linear/equal-power/smoothstep fades. Apply is one Undo/Redo item; save/reopen preserves exact settings. Split/trim retain the source envelope and raw recordings stay unchanged. The shared playback/export reader preserves float headroom. Controls are read-only while audio is prepared; live clip automation, timing/rate/ pitch and refreshed installed packages remain separate work. See the workflow, evidence and remaining gaps.