performance-analyzer

Memory leaks in VoiceLite occur when audio buffers or Whisper processes aren't disposed, causing RAM to grow over time. Use dotMemory or Task Manager to monitor idle RAM against the 150MB target, then profile to locate unreleased byte[] arrays or Process objects in AudioRecorder and PersistentWhisperService components.

Profile VoiceLite using Task Manager for real-time monitoring, dotMemory for heap analysis to detect memory leaks, and PerfView for CPU profiling across components like AudioRecorder and MainWindow. These tools help identify which processes exceed performance targets.

Optimize Whisper transcription speed by comparing model sizes and applying command-line flags to reduce latency. Test different configurations against your transcription latency targets to balance accuracy with performance while keeping the application responsive.

Yes, performance-analyzer monitors idle/active RAM, CPU, and transcription latency against strict requirements you define. It enables threshold-driven alerting and regression testing to ensure VoiceLite maintains non-negotiable performance targets across all components.

VoiceLite targets include idle RAM under 150MB, controlled active RAM and CPU usage, and acceptable transcription latency. These thresholds apply across AudioRecorder, PersistentWhisperService, and MainWindow; exceeding them signals bottlenecks requiring profiling and remediation.

Startup time is a non-negotiable performance target because slow initialization impacts user experience. Performance-analyzer tracks startup latency alongside runtime metrics like RAM and CPU to catch regressions early through automated monitoring and testing.