Implements multi-resolution cross-correlation for synchronizing audio tracks using Accelerate vDSP. Features pyramid coarse-to-fine correlation, consensus clustering across chunks, drift fitting, and confidence thresholds. Self-contained algorithm reusable in any audio synchronization project.

On-device beat and downbeat detection using a bundled Core ML model. Handles chunked inference with border discard, peak picking, and BPM estimation. The detection logic is isolated and could be extracted for other audio analysis applications.

A registry of compositing effects defined by descriptors with parameter specs and apply functions. Provides a clean plugin architecture for adding new effects without modifying the compositor core. Reusable pattern for any application needing extensible effect pipelines.

File-based cache with SHA-256 keys derived from file attributes (size, mtime). Supports memory and disk tiers, stale entry cleanup, and thread-safe access. A reusable caching utility for any Swift application.

A Swift concurrency semaphore that limits concurrent async tasks. Provides a simple API for gating resource-intensive operations like waveform extraction. Reusable in any async Swift project.