Skip to content

M5: 46 audible calibrated presets and compact 612px premium compressor GUI - #8

Merged
masarray merged 10 commits into
mainfrom
milestone/m5-factory-preset-browser
Oct 31, 2026
Merged

masarray merged 10 commits into
mainfrom
milestone/m5-factory-preset-browser

Conversation

@masarray

@masarray masarray commented Oct 31, 2026 •

Copy link
Copy Markdown
Owner

M5 Reconciled — 46 audible factory presets + narrower premium GUI

This PR consolidates the user's two requested improvements on top of the existing 46-preset searchable/favorites browser; it also retains the recent fix that preserves the displayed preset name when opening/dismissing the browser.

Preset sound: stronger first-use impact, controlled gain

  • All 46 factory programs now have a nonzero, intentionally calculated initial Makeup based on the compressor's own knee/range/ratio curve at a nominal −18 dBFS peak signal plus a small audible starting lift. Calibration is performed once when the static factory catalog initializes, never during audio processing.
  • Makeup is use-case bounded: Mastering ≤3.2 dB; Vocals/Live ≤6 dB; other groups ≤5.8 dB; special effects ≤8 dB. Parallel Mix scales the estimated gain reduction contribution.
  • “Neutral Start” now actually compresses (−24 dB threshold, 2.5:1, moderate attack/release/knee) rather than preserving a near-zero-reduction factory state.
  • The estimated boost is a preset starting point. This is not active loudness normalization, a limiter, or proof of audible improvement for every arbitrary input. Loud inputs may clip in later devices. Adjust makeup, input, monitor OUT, then loudness-match bypass for serious comparisons.
  • Add 46-program regression covering ranges, unique IDs, safe live routing, nonzero makeup, gain reduction at a −12 dBFS reference and finite, sub-0 dBFS output on nominal test material. Preserve state schema, live routing switches, VST3 parameter IDs and zero additional DSP samples.

Much narrower, still premium UI

  • Full default 704×520 → 612×500 (15% smaller window area); minimum 660×490 → 580×480. MINI 320 px tall.
  • Single-row compact header with small but legible brand/model/preset/A-B/FILE/bypass, browser overlay retains 46 programs with search/categories/favorites.
  • Main knobs shrink from 70px to 62px control height, keeping full mouse hit area, captions and the brushed-metal style.
  • SIDECHAIN uses a strict fractional width grid: HP/LP tracks + detector + external + listen stay inside 580px without clipping.
  • Keep isolated opaque LCD and UI event coalescing. No change to DSP/audio callback/driver.
  • Update header geometry/knob-size/MINI/full assertions in Windows smoke, plus factory test and engineering docs.

CI and actual-device gates

Source SHA 7073098ba120b988e8b2c24ce477d14bc3ff99aa; check exact SHA Windows Release VST3+Standalone, plugin_smoke UI/preset bank and Linux engine-safety. Do not merge without both jobs passing. Then publish Windows artifact with default/minimum/SIDECHAIN/MINI/browser screenshots and SHA256. Hardware listening/headroom, PS22 stability and long-run sessions remain separate user acceptance gates.

FabFilter Pro-C 3 publicly explains wet makeup/auto-gain as compensation and warns its auto-gain is an educated estimate, not measured loudness. Reference: https://www.fabfilter.com/help/pro-c/using/timecontrols . All ArSonComp presets remain original; none of FabFilter's .ffp data, assets or proprietary compression DSP are reused.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 31, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-31T15:35:37.772605Z 776745c PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 776745c661

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread plugin/Editor.cpp
Comment on lines +638 to +640
for(std::size_t i=0;i<bank.size();++i)
if(favorites[i])data+=juce::String(bank[i].id)+"\n";
file.replaceWithText(data);

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Preserve favorite changes made by other plugin instances

When two plugin editors are open, each loads favorites only during construction and subsequently replaces the same application-wide file with its private snapshot. For example, opening editors A and B, starring one preset in A, and then starring another in B silently removes A's favorite from disk; unstarring can similarly resurrect favorites removed elsewhere. Apply each toggle against the latest shared state, or synchronize favorites across instances, while keeping persistence on the message thread.

AGENTS.md reference: AGENTS.md:L27-L27

Useful? React with 👍 / 👎.

@masarray masarray changed the title M5: 46 original factory compressor presets with searchable category browser and favorites M5: 46 audible calibrated presets and compact 612px premium compressor GUI Oct 31, 2026

Copy link
Copy Markdown
Owner Author

M5 reconciliation audit — 31 Oct 2026 (evidence-based)

Branch gate: PR #8 is still OPEN and NOT MERGED. Main is M4 (704x520 FULL, 660x490 min, 330 MINI) and has only 6 factory presets in plugin/Editor.cpp. Do not describe M5 as shipped on main.

Source verified on M5: plugin/HeaderLayout.h configures 612x500 FULL, 580x480 min, 320 MINI; the 7 primary controls use 62px height and the sidechain grid is width-relative. plugin/FactoryPresetBank.h defines 46 original programs / 10 categories, deterministic reference-level Makeup Gain with per-category caps, and preserved live routing switches. Search/description/favorites are implemented in the transient browser, not yet confirmed by physical UX test. These are source-code facts, not final visual or sonic acceptance.

Prior CI failure: run #51 (41219048907) built VST3+Standalone and passed Linux engine-safety and Windows engine_contract, but Windows plugin_state_editor failed PRESET_BROWSER_80 visible=0 before packaging. Browser UI/screenshot was not validated. Windows source-level preset tests preceding this gate were reached successfully, including catalog count, bounds, nominal DSP finite outputs and nonzero makeup. Run #50 also failed.

Bug fix reconciliation: commit 08c1466 handled offscreen synchronous open; commit 61a4c19 fixes the Windows smoke's repeated terminated JUCE event-loop usage by exercising the real asynchronous ComboBox callback in the FIRST message loop. Neither changes DSP or host parameters. CI #53 (41219778809) is the relevant exact-head check for 61a4c19 (status pending at writing).

Acceptance gates (must not collapse into a single "done" claim)

  1. Windows Release compile VST3+Standalone, plugin_state_editor and 5 screenshots (default/minimum/SIDECHAIN/MINI/browser), Linux sanitizer: all PASS on exact M5 head.
  2. Open and visually inspect all compiled screenshots for text clipping, collision, readable 580px header/controls, browser overlay and sidechain. Verify actual program selection and browser search/favorites/dismiss in a Windows desktop/DAW.
  3. Confirm all 46 presets audible/useful on representative vocals/drums/bass/mixbus material with loudness-matched bypass, track GR, crest/peaks and clipping, no assumption that louder automatically means better. Existing -18 dBFS static makeup compensation is not real-world normalization or true-peak limiting.
  4. Check a real PS22 at 48kHz ASIO (512, 256, 128 samples) for startup/persistent crackles, XRUN/dropouts, DPC/USB and host-bypass A/B. Earlier M4 ASIO backend enablement is not proof of a crackle-free device. The M5 UI/preset work does not change compressor DSP.
  5. Only after evidence passes, merge M5; keep physical/sonic acceptance explicitly pending if not performed. No FabFilter Pro-C 3 algorithmic parity claim.

Reference: PR #8 and CI runs #50/#51/#53. Avoid repeating already merged M2–M4 work or creating unnecessary CI runs.

@masarray
masarray merged commit cbb2a33 into main Oct 31, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant