admin
2c96a223af
feat: CC11 expression automation (spec midi_note_cc11_expression)
...
- MIDINote schema + expression_curve (position_ratio 0..1, value 0..127)
- Piano roll CC lane mode Expression (CC11): draw/move/del anchors
- Offline: _expression_events grid 64 samples -> job.events -> RenderJob.cpp sample-accurate CC
- Realtime: scheduleNoteExpressionCC11 grid 16ms via bridge pushEvent / SonicSF fallback, epoch-guarded, cleared on stop
- tests/test_cc11_expression.py 6/6 (unit interp/grid/clamp + integration energy follows curve)
- rebuild daw_vst_bridge (narrowing fix) + sync binaries/dist, babel rebuild + sync 5 copies
2026-08-26 11:14:56 +07:00
admin
4932bb88db
fix(bridge): restore VST2 preset chunk on respawn + offline render (Qin silent fix)
2026-08-20 10:18:18 +07:00
admin
7c22ae1f02
refactor(bridge): split fx bridge, stabilize VST3 host chain, drop old install engines
...
- remove diagnostic canaries from audio hot path (crackle fix) and
stuck C5 note raw-MIDI test (Vst2Instrument::init noteOff)
- split native bridge into daw_vst_bridge (VSTi) + fx_vst_bridge (VST FX)
- VST3: processData prepare before setActive, zero input buses, accept
both .vstpreset and bridge-state blob in RenderJob presets
- native_render/vst_engine: bridge-exe discovery + render_fx_chain + /fx-gui
- delete old install/ engine binaries (install via sidecar+resources now)
- add BUILD_LAYOUT.md, build_selfcheck, tools/ diagnostics
2026-08-20 09:42:42 +07:00
3dtours
ba9ceeea90
fix(bridge): WAV header data_size for IEEE float32 (was 16-bit byte count)
...
- finishWav call sites accumulated n*4 (stereo 16-bit) but float32 writer
emits n*8 bytes -> RIFF/data chunk sizes half the real value; players and
soundfile truncated renders to half length
- gate test now asserts file size == 44 + data_size to catch this
2026-08-18 17:35:54 +07:00
3dtours
a2f93138be
Normalize Linux audio pipeline to lossless Float32 (skill: lossless-audio-compliance)
...
- native bridge WAV writers: 16-bit PCM truncation -> IEEE Float32 (fmt 3, bit-exact)
- vendor sheredom_json.h (vst3sdk 3.8.1 lacks moduleinfo/json.h)
- Dockerfile: drop nonexistent sfizz tag v1.2.3
- Python sf.write -> subtype=FLOAT everywhere
- resample_poly -> kaiser(16) window (>=130dB stopband)
- audio_editor._write_wav_lossless(): FLOAT passthrough, TPDF dither only for
8/16/24-bit final exports
- app.jsx WAV encoders: Float32 default, TPDF dither for integer depths, fix 24-bit
- tests: lossless compliance gate (null test < -140dBFS, float32 format) 9/9 pass
- TASKS_WINDOWS_AUDIO_NORMALIZE.md: Windows-side remaining tasks
2026-08-18 11:50:26 +07:00
admin
e79dae5d5c
phase4: Linux native_bridge build (Docker) — sfizz from source, --render SF2/SFZ
...
- Dockerfile: build sfizz tu native_bridge/sfizz-src (hoac git clone v1.2.3),
sed fix sfizz.pc (-llibsfizz -> -lsfizz), cmake bridge voi 3 flag SMTG OFF
(VSTGUI/plugin/hosting examples) — bridge headless chi can --render.
- CMakeLists: nhanh non-WIN32 build daw_vst_bridge tu main_linux.cpp +
RenderJob.cpp + NativeInstrumentEngine.cpp (SF2/SF3 + SFZ; VST3/VST2
hosting Windows-only, guard #ifdef _WIN32). gui_probe/plugin_host/
seh_channel_test chi build tren Windows.
- RenderJob.cpp: JSON parser (sheredom json.h) khong con phu thuoc
HAVE_VST3SDK — render SF2/SFZ chay duoc khong can VST3 SDK.
- NativeInstrumentEngine.cpp: guard windows.h/VST3/VST2/Sandbox/SEH.
- main.cpp: g_engine unconditional + guard disable_ime_contexts (_WIN32).
- engine.spec: bo pedalboard collect_dynamic_libs + hidden imports.
- vst_engine/render_engine: DecentSampler -> NotImplementedError + huong
dan convert .dspreset -> .vstpreset; pianobook fallback synth.
- Verified: WSL Ubuntu 24.04 build OK, bridge Linux --render SF2 ra WAV
44100 stereo peak 1575; Python tests 115 passed 1 skipped.
- Dockerfile build chua test duoc (may khong co Docker).
2026-08-17 11:46:24 +07:00
admin
bbb11f8066
Phase 1: import .vstpreset in offline render (loadSerializedState)
...
daw_vst_bridge --render now applies a VST3 preset when job has 'preset'
(base64 of a .vstpreset file). Parse the SDK binary chunk format with
PresetFile (vst3sdk), rewrap component/controller state into the bridge
blob [4B BE compLen][comp][4B BE ctrlLen][ctrl] and apply via
loadSerializedState before rendering. SF2/SFZ ignore preset (warning).
Bad base64/magic/chunks fail the job with rc 1 and no partial clip.
Tests: golden Nexus .vstpreset fixture; VST3 with/without preset,
bad preset -> rc 1, non-VST3 preset ignored. 20/20 pass.
2026-08-17 10:44:13 +07:00
admin
0cca84c5f9
feat(bridge): add --render offline CLI (Phase 0 of pedalboard replacement)
...
daw_vst_bridge --render <job.json> --out <clip.wav> renders MIDI notes
through the same instrument engines as realtime (SF2/SFZ/VST3/VST2) to a
16-bit stereo WAV, headless, no SHM. Mirrors plugins.py/vst_engine.py
note semantics (velocity int(vel*127), bank CC0 + program at t=0, beatSec
60/max(30,bpm), totalNeeded max(1024,endSample), no tail). Chunk clamped
to 256 to match VST3 host prepared buffers. Uses vendored sheredom json.h
(gated on HAVE_VST3SDK); rc: 0 ok, 1 arg/job error, 2 load fail, 3 crash.
Partial output removed on failure. Verified: SF2, SFZ, Nexus VST3 render
non-silent at exact duration; 9/9 tests pass.
2026-08-17 08:04:25 +07:00