Repository navigation
Conversation
…esponse and visual QA
…ick EQ and dual output meters
…ts, undo/redo and context menu
…ynamic bell/shelf
1 of 8 tasks
…ector and detector retune
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
M0 → M2 integrated ArSonEQ VST3
Native engine and audio quality
M2 MaSonKuPik-inspired native-web GUI (requested direction)
masarray/masonkupik/src/components/plugins/EQEditor.tsxand_shell/spectrumCanvas.ts; port interaction/visual language, not its React/mastering runtime or Web Audio.ARSONEQ_USE_NATIVE_GUI=1forces fallback when debugging a WebView2 host.Evidence at commit d579962
Important: the lightweight asset bundle does not make Chromium WebView2 itself free in RAM; benchmark before claiming an ultra-light editor. This PR remains Draft/test-only until DAW validation and stability gates are satisfied.
M2.1 full-canvas UX correction — user screenshot review (2026-10-29)
masarray/arsonmix/design-references/brushed-metal-v1.0/reference.json; aluminum controls, dark blue-black FX glass, cyan activity, colored filter nodes. No FabFilter visual assets copied.M2.1 validation
b478cbbf380fc0218f25e199f75108a0dcce5c04).CI policy: final DSP workflow tracks WebUI changes, while Windows VST3 tracks actual embedded asset changes but not documentation-only/smoke-test-only changes. Avoid rebuilds for small editorial amendments.
Final test-binary provenance: Windows plugin binary from Actions artifact ID 12909735983 (CI's historical artifact label still reads M1.5; the actual source SHA is
b478cbbwith the complete M2.1 GUI embedded). Windows binary SHA-256:810da92d0dcf74cda641bc561e0217dcc8aa13e0eb8c161dc041a65eea779ea1, size 9,148,416 bytes. Linux UI smoke, 3 DSP tests; Windows compile + 3 DSP tests passed. Draft retained until actual DAW visual/load/recall profiling.M2.2 — BOTH real Windows deliverables (2026-10-29)
FORMATS VST3 Standalonebuilt from the same native C++ DSP and M2.1 WebView2 editor, not a browser-only standalone page. PR branch headc6793581b8b0cb8d1973ed9c9fc402feb771eb3a.ArSonEQ-M2-2-Windows-x64-VST3-Standalone-TEST. Downloaded, inspected archive, ZIP CRC integrity PASS, both PE MZ binaries and embedded WebView2 UI strings verified.Standalone/ArSonEQ.exe: 9,517,056 bytes; SHA-2563e167c368e1ed4f2186ac5c5bdbf3b4a65ffee9432b225a7cdc4886f96c01479.VST3/ArSonEQ.vst3/Contents/x86_64-win/ArSonEQ.vst3: 9,148,416 bytes; SHA-256764268c31e3d1b526492177207cde67b0db50b1e2c6f1280261148fb2f069a97.BUILD.txtavailable in artifact. Note: BUILD.txt's commitd6f9a5a...is the GitHub PR merge test SHA used by checkout, not the raw PR branch head.Release status remains Draft / ENGINEERING TEST. Building both PE files is now verified; actual device audio and DAW operation need physical Windows testing.
M2.3: VSTHost-confirmed occlusion bug / flagship UX redesign — commit ce164aa
Real screenshot evidence: ArSonEQ M2.2 card obscures the response curve and node at approximately 78Hz/+12dB. Pro-Q 3 comparison shows a more compact, grouped instrument panel near the bottom and a dominant unobstructed frequency response.
Structural fix (not CSS repositioning): 143px selected-band control deck is now a flex sibling outside the graph, after
</div id=graphWrap>. The chart dynamically receives remaining height and can never be hidden underneath a band editing card, including when resized or placed near extreme frequency/gain. Deselection removes the deck and returns full chart height.Premium visual pass: warm gold native C++ EQ curve, green boost shading and subtle negative-gain shading, charcoal/muted violet EQ plot, narrow separate FFT dB rail, selected-band colored response, small metallic knobs and ArSonMix-control material language. Right FFT rail is excluded from frequency X math; dB scale defaults to ±12 and expands to ±24 if required.
Performance & compatibility: Canvas 2D remains event-driven; FFT only on editor UI thread and gated by Analyzer; no DSP/host schema changes; same assets feed both VST3 and Standalone. Replaced rather than appended bloated legacy CSS, maintained offline/zero-React packaging. A dedicated synthetic CI preview is NOT part of production binaries or DSP.
Verification plan:
Keep Draft until both artifact and actual host screenshots are verified.
M2.3 validation evidence and final visual-polish commit (2026-10-29)
ce164aacompiled both Windows x64 Release targets successfully, including 3/3 DSP Windows tests: https://github.com/masarray/arson-eq/actions/runs/40906386424a354f44passed Linux DSP 3/3, JS fake-host smoke and 900x535 headless Chromium screenshot: https://github.com/masarray/arson-eq/actions/runs/40907769706. The screenshot was inspected: graph and selected band remain visible, the dock is physically below the graph, and upper/lower gain axis labels are no longer stacked. It also passed both Windows Release formats and 3/3 DSP Windows tests: https://github.com/masarray/arson-eq/actions/runs/4090776976630c0308moves the upper HUD safe right margin from 48px to 105px, preventing text intrusion into the dB axis region.No DSP or host parameter code changed in M2.3. Audio RT tests continue to guard the preexisting stable engine. Do not declare Pro-Q 3 parity; this milestone specifically improves professional visual hierarchy and the user-reported curve occlusion bug.
Verified final M2.3 binaries (engineering test; PR head 30c0308)
ArSonEQ-M2-2-Windows-x64-VST3-Standalone-TEST; the CI artifact label is historically M2.2 but payload contains the M2.3 UI).Standalone/ArSonEQ.exe9,519,104 B, PE MZ and M2.3 embedded WebUI; SHA2563dff30f0bf1c2354db74c8406c7afb727e4bf6fd91de520071dcd393158c00ba.VST3/ArSonEQ.vst3/Contents/x86_64-win/ArSonEQ.vst39,150,976 B, PE MZ and same WebUI; SHA256a9c2dcdad93ad619281dc500c5717dec575eeef71b44794ac50508db13709365.86cb0f815904d90b3dbdf68abb9cf8287559cf4c; branch HEAD with final CSS change is30c030876a1d03c8fd73f6d2d32bfdd1c95892b7.Remaining blockers: real VSTHost and ArSonMix GUI inspection on Windows, Standalone soundcard input/output routing test, 100%/150% DPI and plugin memory/CPU profiling, host validator and automation/state recall. Keep Draft; do not claim parity with commercial Pro-Q 3.
M2.4 strategy approved as engineering baseline — 2026-10-30
Context: After a real user screenshot comparison of ArSonEQ Standalone (902×563) and FabFilter Pro-Q 3 running in VSTHost, audited ArSonEQ DSP, pre/post analyzer, WebView2 canvas and JUCE wrappers against FabFilter's 50-page Pro-Q 3 manual.
Major root causes: (1) analyzer source sums stereo
0.5*(L+R)before FFT, incorrectly cancelling anti-phase material; (2) 4096 FFT nearest-bin mapping to 224 log x points and duplicate smoothing state, which contributes to stair steps; (3) no calibrated 4.5dB/oct spectrum tilt, display speed/resolution/range/freeze or output meter; (4) selected-band dock ~148px leaves only ~320px of graph at 900×535; (5) 6 original EQ types and 12–48dB cutoff slope are not FabFilter's 9 shapes and 6–96dB/brickwall/full processing modes.Milestone M2.4: fix stereo energy telemetry and FFT/log-bin aggregation/normalization first, add optional 1024/2048/4096/8192 FFT, temporal attack/release, 4.5dB/oct visual-only tilt, pre/post/freeze and sample-peak meter. Shrink selected quick editor to <=70px (chart >=390px at default size), gate advanced controls. Add anti-phase L/R, frequency/amplitude calibration, equal-output test, latency/CPU and host screenshot regression gates. Use one consolidated Windows VST3+Standalone build, not cosmetic 1-line CI cycles.
Next stages: M2.5 advanced direct band editing; M3 dynamic EQ and verified detector; M4 phase/algorithm accuracy and new filter shapes; M5 spectrum references/EQ match. Do not claim Pro-Q 3 sound or feature parity from UI screenshots.
Documentation committed (no binaries changed):
docs/PROQ3_GAP_AUDIT_AND_M24_PLAN_20261030.md(primary gap matrix, numerical tests, no-occlusion UX acceptance, 5 subsequent phases).docs/ARD.md,docs/PRD.md,docs/ROADMAP.md,README.mdupdated.c6a81c1d782e8af94afbf1111a53958589dc152fis documentation-only, so no new Windows CI is needed or claimed. The earlier M2.3 engineering binaries remain test-only and do not include M2.4 code.M2.4 native FFT + compact graph implementation (2026-10-30)
Commit:
9bfa3268bdc8241404023a38506506e8a1fb4f55— 16 changed/added files.0.5*(L+R)analyzer false cancellation; four pre/post L/R PCM streams in non-blocking SPSC ring, mono compatibility.PluginEditor.cpplegacy telemetry field references; fixed, including proper stereo-native fallback and FFT Hann-window precomputation, in9b0127ab3a9c536356c705bbbac0d0928a9404dc.Implementation note: the four-channel power algorithm prevents anti-phase cancellation; low-frequency FFT resolution remains constrained by FFT size and sample rate. The 900x535 screenshot uses a synthetic UI fixture for layout only, never published as native audio proof. Analyzer menu settings currently live in the editor session, not host automation. PR remains DRAFT until physical host verification.
Final M2.4 artifact provenance (verified 2026-10-30)
9b0127ab3a9c536356c705bbbac0d0928a9404dc.40966762003SUCCESS: bothArSonEQ_VST3andArSonEQ_Standalonecompiled, all 3 native DSP suites PASS, packaging passed.12939505028,ArSonEQ-M2-4-Windows-x64-VST3-Standalone-TEST; ZIP integritytestzip() is None, modules verified MZ/PE.Standalone/ArSonEQ.exe, 9,524,736 B, SHA25654948de214a324005fb568f1edc06775ffe03e456f7d97f29a3be00dcfebc984.VST3/ArSonEQ.vst3/Contents/x86_64-win/ArSonEQ.vst3, 9,155,584 B, SHA256d9ddf7bdb95ac48afa432f3d3cc4e9f6d04624dd944ef5f89545c17feb3a3861.Commit: 30ecbb1324470f722459d2b6d546eb5b5b2a57dfis GitHub Actions PR merge-test SHA used by checkout; the branch HEAD above is the actual source revision under review.README_WINDOWS_TEST.txt,BUILD.txtand native x64 VST3 bundle + Standalone EXE.M2.5 professional UX milestone — screenshot regression FIX (2026-10-30)
Source commit
751316a18a532cd2156ab561e239461cb588e13f, 12 files. User M2.4 Standalone screenshots showed band panel causing vertical graph reflow when clicking/releasing dot. Root cause wasinspectoras a flex sibling; selection toggled display height by 70px and MORE by ~151px, changing the canvas pointer coordinate system.Implemented (real WebView2/C++ bridge, not just design):
bulkEditIPC (<=168 field changes) andgestureMany(<=24 bands), per-band host automation gesture opening/closing; no audio-thread changes.Not yet implemented / don't overclaim: FabFilter Pro-Q 3 parity, dynamic EQ, Natural/Linear Phase, true peak, band audition/solo, DAW-global undo, EQ Match and reliable host-validated CPU benchmarks. PR intentionally remains Draft.
Final M2.5 engineering artifact verification (2026-10-30)
12940093007(ArSonEQ-M2-5-Windows-x64-VST3-Standalone-TEST), generated from same PR branch revision751316a18a532cd2156ab561e239461cb588e13f.testzip() == None, valid MZ/PE headers on both outputs.Standalone/ArSonEQ.exe9,536,000 bytes, SHA256a8cf99f63153fc606c7475f96260582b31e0ee44b7bedb2b81da94b1c67e2278.VST3/ArSonEQ.vst3/Contents/x86_64-win/ArSonEQ.vst39,167,872 bytes, SHA256fa9d9bdbc293752642f674c47e3124c9be9553d4ee9f9ea173fd550b9b1ae7d8.ArSonEQ-M2.5-900x535.pngchecked manually: selected band card is over bottom graph, full-height EQ plot and nodes remain present; synthetic curves only for QA.M3 native presets and first dynamic DSP (integrated 2026-10-30)
Commit
f304712fc0a49f8a7f636d34f35dbebd31d49837, one integrated 17-file commit. User-confirmed M2.5 floating inspector retained.Do not overclaim: advanced auto-threshold, program-dependent knee, external sidechain, linear/natural phase and EQ Match are not implemented. Their independent plans live in #2 (M4) and #3 (M5). The current M3 dynamic filter is an original lightweight manual-threshold algorithm, not a copy or sonic equivalent of FabFilter Pro-Q 3. The UI plot is static steady-state response; dynamic response ghost curve still future.
M3 final CI engineering-test provenance
f304712fc0a49f8a7f636d34f35dbebd31d49837, test runner commit may be PR merge SHA.ArSonEQ-M3-Windows-x64-VST3-Standalone-TEST. ZIP archive CRC check OK.Standalone/ArSonEQ.exe9,559,552 bytes, SHA256dc5a0b3827334ce2d645db75135c16b20652d77d4dbb095f8b57b2fc9c1885a7.VST3/ArSonEQ.vst3/Contents/x86_64-win/ArSonEQ.vst39,190,400 bytes, SHA2569aad70a18229de39401aa0670353e3c6ad19b2a941732270ceae0473092c7552.M3.1 screenshot-driven fixes + live native dynamics (2026-10-30)
Commit
ed355eb5ede44fa9a5f4c7dba913707e80191aafaddresses latest user screenshots:eq-dynamicswith 256-point L/R curves, per-band effective center gain and activity. UI displays moving gold response, subdued static curve, tinted difference region, thin moving per-band marker, numerical dynamic gain and activity meter. No fake signal animation. No DSP/JS/FFT in processBlock.Correctness boundary: UI dynamic gain/transfer is derived from the current native DSP coefficient model and measured wet factor; it is an instantaneous stationary approximation of a time-varying dynamic IIR, not a directly sampled transfer function. Does not implement Pro-Q 3's proprietary analog matching/auto threshold/program-dependent ballistics, nor iZotope external detector routing. M4/M5 still open.