Skip to content

experiment(audio): reversible Spatial Audio quality, DSP safety and regression suite - #4

Merged
Batmancpu merged 8 commits into
mainfrom
experiment/temporary-spatial-audio-quality-20261114
Nov 14, 2026
Merged

Batmancpu merged 8 commits into
mainfrom
experiment/temporary-spatial-audio-quality-20261114

Conversation

@Batmancpu

Copy link
Copy Markdown
Owner

Temporary, isolated experiment

  • Maintains the existing FC Spatial Audio source unchanged. Adds an opt-in build-only fc/spatial-audio/experimental/apply.py and config.properties flag.
  • Reduces overly aggressive widening/crossfeed, uses a sample-rate-specific low-pass, adds a common-stereo-channel sample-peak ceiling (0.98), 250ms limiter release, and a 12ms crossfeed delay.
  • Removes PCM16 ByteBuffer duplication and per-frame delay-ring modulo; dispatches rare completed fingerprints to one bounded background worker.
  • Raises Auto learning confidence (80% On, 85% Off) and independent-evidence requirement (3.5) while keeping all existing overrides, Atmos, speaker and DSP bypass protections.
  • Adds 11 regression tests for sample-rate variations, bit-exact disabled operation, mono bypass, PCM16, clipping, toggles, background callbacks and Auto learning.
  • Adds pull-request-specific Android test workflow that runs the FC identity, original FC Spatial Audio, experimental layer, brand UI overlays, locale fallback and production unit tests without signing or releasing.
  • Versioned release workflow triggers on toggling experimental flag and restores the source changes before upstream sync commits.

Rollback

Flip enabled=false in fc/spatial-audio/experimental/config.properties and merge. The next successful workflow releases the original stable FC engine, with an incremented Android versionCode, unchanged package ID, signature, brrMusic branding and Fluid Cloud behavior.

Validation gate

Do not ship before the isolated production test workflow completes. A passing CI job verifies algorithmic regression tests and compilation, but a device-listening A/B test is still needed before claiming superior subjective sound quality. Source diff must not include modified preexisting fc/spatial-audio/SpatialAudioProcessor.kt or SpatialAudioLearning.kt.

@Batmancpu
Batmancpu merged commit 5e2e979 into main Nov 14, 2026
1 check 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