Problem: System was stuck using single all_tracks folder, causing: - No bucket sampling diversity (all files in 1 folder) - Repetitive sample selection - No coherence between sections - Fades/volumes not auto-applied Fixes: 1. Changed DEFAULT_LIBRARY from all_tracks to organized_samples - server.py: Updated SAMPLES_DIR - sample_manager.py: Updated base_dir - health_check.py: Added organized_samples as primary paths 2. organized_samples structure enables T013 bucket sampling: - loops/bass: 34 samples - loops/synth: 43 samples - loops/vocal: 24 samples - oneshots/kick: 20 samples - oneshots/perc: 35 samples - Each subfolder < 15 files = perfect for bucket sampling 3. Added auto-automation to generate_song(): - Fade-in 4 bars for kick/bass/hat (intro) - Build curve: music tracks 0.5 -> 0.9 (32-40 bars) - Reverb automation: 0% -> 40% -> 0% on atmos/pad/vocal - apply_automation parameter (default True) 4. Each track now gets diverse samples from different subfolders: - Bass from loops/bass (34 options) - Synth from loops/synth (43 options) - Drums from oneshots/ (kick, perc, snare) Coverage wheel will now track usage across 20+ subfolders instead of 1. Diversity memory will work correctly with proper family tracking per folder. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2.1 KiB
2.1 KiB