Skip to content

V3.3 Control Usability

Choose a tag to compare

@dfeen87 dfeen87 released this 24 Oct 17:23
· 227 commits to main since this release
d182f45

ISLA v3.3.0 — LIVE Control & Reliability

Release preparation · 24 October 2026 · eight merged post-v3.2.0 pull requests

Scope: User-facing LIVE control usability, reliable Qt mode selection, terminal guidance, safe audio-device selection, physical listening-output attenuation, and improved macOS Audio MIDI documentation. This release does not introduce a new harmonic algorithm or tune frozen scientific experiments.

Immutable baseline: v3.2.0 — Audio Control Center, commit fee185c5f1851c6b7fce547068d86b756ea12332, published October 24, 2026. The v3.3.0 release candidate includes the following subsequent merged PRs: #38, #39, #40, #41, #42, #43, #44, and #45. There are no unmerged PR changes intentionally included.

Publication status: This is the source release note, not proof of a published GitHub release. The v3.3.0 tag/release must be created separately from the final approved and green main commit. Do not publish this note as a claim of native physical-Mac certification or unattended playback reliability.

Highlights

1. Reliable LIVE effect selection and confirmations — #39, #41, #43, #45

  • LIVE starts in Original / Bypass in the graphical Control Center; effect changes select Velvet, Legato, or Melt without restarting the running audio stream.
  • Step 4 now uses a single accessible effect dropdown and a separate, explicit Reset to Original action. Re-selecting the same effect is idempotent, rather than toggling unpredictably.
  • Corrected a user-reported bug where periodic state refreshes temporarily disabled the dropdown, dismissed its popup, or reverted a newly selected mode to Bypass. The GUI now preserves pending intent, reconciles asynchronous success/failure, rejects stale snapshots and distinguishes requested, confirmed, and DSP-applied mode.
  • Regression coverage now uses real Qt popup mouse/keyboard interactions in addition to model/worker state checks; operator verification on real Mac hardware remains separate.

2. Dedicated LIVE Listening Volume — #41

  • New 0–100% GUI listening-volume control with Apply Volume and Mute/Unmute, restoring the previously selected nonzero setting when unmuted.
  • Applies attenuation only to ISLA's post-DSP physical output, not BlackHole capture, the system's master output, local ALIVE files or other apps. 100% is unity (no additional gain), and 0% mutes.
  • Uses a short smoothing ramp (approximately 5 ms) in the existing worker path; changing the level does not restart LIVE. It does not provide universal loudness normalization or establish musical superiority of any effect.

3. Responsive Stop, safer device polling and diagnostics — #39, #40

  • Avoids routine full CoreAudio/PortAudio enumeration while streaming, preventing UI polling from competing with Stop and mode commands. Stop releases audio before any subsequent optional recheck.
  • Surfaces the actual LIVE fail-closed safety-stop reason near Step 4 while preserving detailed metrics and error counters.
  • Enlarged, scrollable diagnostics retain the reader's scroll position during health updates, while selected/active effects and processing acknowledgment are visible.

4. Numbered, verified capture/output pickers — #43

  • In isla>, after source system, typing bare input or output now opens a filtered numeric picker. Type only the displayed number at its selection prompt; cancel exits without mutation.
  • The LIVE input picker allows the standalone BlackHole 2ch capture device; the output picker allows a separately verified physical stereo device, never the BlackHole feedback sink or an input-only microphone.
  • PortAudio index and CoreAudio-native identity are rechecked before applying a picker choice, guarding against hot-plugged/stale selections. Prior setup audio flow, explicit numeric selectors and exact-name commands remain available.
  • Device selection and rescans remain unavailable while a LIVE stream is active. ISLA does not change the Mac's default output automatically.

5. Contextual, professional terminal command center — #45

  • A width-aware, ASCII-friendly DOS-inspired help screen shows the current source, running/stopped state, mode, devices, relevant commands and actionable next step.
  • Topic help includes help live, help file, help devices, help modes, and help troubleshoot, plus ?.
  • Mistaken play during LIVE, repeated monitor start, and attempted output changes during active monitoring now provide contextual, safe instructions. These command responses do not reopen or hijack a live stream.
  • The existing diagnostic status output, file player commands, and shutdown behavior remain available.

6. Safer and clearer installation/onboarding — #38, #42, #44

  • Repaired and embedded the real ISLA Control Center screenshot in the README.
  • Added prominent warnings to lower source and physical listening volume before using any 100% output command. AppleScript output-volume commands affect whichever macOS output supports them; they are not dedicated BlackHole gain controls.
  • Added two authentic Audio MIDI Setup photographs demonstrating BlackHole 2ch at 1.0 / 0.0 dB unity and a deliberately low MacBook speaker volume with centered balance. These are device-specific examples, not a promise that the speaker screenshot equals 25% macOS output.
  • Documented independent macOS alert-volume adjustment (e.g. osascript -e 'set volume alert volume 25') and Sound Effects routing.

Compatibility and commands

  • Python: 3.10+; distribution isla-audio, import isla, with compatible legacy phased_music_notes imports.
  • Desktop UI: optional PySide6-Essentials, installed using python -m pip install -e '.[gui]'.
  • Launch: isla or python -m isla for terminal; isla window or python -m isla window for Control Center; window from an active terminal session.
  • LIVE: experimental macOS 13+; standalone BlackHole 2ch capture, separate supported physical stereo output and manually configured macOS system routing, typically 44.1 kHz or 48 kHz.
  • ALIVE: local WAV playback and original offline export remain separate from LIVE.
  • Public package, HTTP app and JUCE/CMake version metadata advance together to 3.3.0. No VST/AudioUnit certification, standalone Mac installer or production professional real-time guarantees are introduced.

Example LIVE workflow (device numbers depend on the current machine):

isla> source system
isla> input
Select input: 0
isla> output
Select output: 1
isla> monitor start
isla> mode velvet
isla> bypass
isla> monitor stop

Within the GUI, Start LIVE and Stop LIVE remain explicit operations; changing an effect or ISLA listening volume does not replace these controls.

Test evidence and remaining acceptance

The merged #45 implementation summary reports 559 passed with 1 existing expected failure for its full local Linux Python 3.12 / optional-Qt suite, and 90 focused terminal/Qt checks. These are reported results on that PR, not automatically evidence that a later release-version commit is green.

Release gates to verify against the exact final v3.3.0 candidate:

  1. Full ALIVE Python acceptance across the repository's supported Ubuntu/macOS CI matrix.
  2. Control Center Qt offscreen GUI acceptance across the supported matrix, including the actual popup and stale-state regression tests.
  3. Linux/JUCE strict-build CI, version/identity tests, and all research preservation checks.
  4. Explicit local/native macOS operator acceptance at supported rates: BlackHole-to-physical-output routing, no feedback, distinct Bypass/Velvet/Legato/Melt selection, pending-state persistence, Reset to Original, listening-volume and mute behavior, Stop/Close, and safe return of the macOS output setting.
  5. Verify package/API/CMake/research identity at the approved exact release commit, and preserve all historical v3.0.0, v3.1.0, v3.1.1 and v3.2.0 tags unchanged.

Limitations: Hosted offscreen GUI tests and fake-audio regressions are not native physical macOS hardware, headphone listening or sustained hot-plug validation. LIVE may incur noticeable latency, underflows or safe-stop interruptions and remains experimental. Stronger native proof, broad device compatibility, sustained-session performance, independent sound-quality assessment and distinct remastered DSP are not claimed.

Research and DSP preservation

  • No new Velvet/Legato/Melt mathematical tuning, normalization or spectral-quality claim is made by this release.
  • The existing streaming audio engine, bounded capture-recovery behavior, fail-closed routing rules, legacy ALIVE workflow, and fixed Experiment 02 research corpus remain under their existing provenance and validation limitations.
  • This version is a control-path and usability release, not a declaration of v4.0 Remastered or an independently validated acoustic improvement.