diodiogod
90a6922903
Version 4.5.25
Optimize Silent Speech Analyzer cache and preview system
Implement two-tier cache system:
- MediaPipe analysis cache (expensive, video-dependent)
- Filtered results cache (cheap, parameter-dependent)
Preview generation improvements:
- Preview now reflects filtered segments when merge_threshold changes
- No expensive MediaPipe re-analysis for post-processing parameter changes
- Reuses cached frame data and landmarks for pixel-perfect accuracy
- Separated preview generation from analysis phase for better performance