- Refined Best Quality alignment across Small, Base, and Large depth models - Added file-based photo import alongside Photos - Added file-based batch import - Fixed video orientation for portrait and rotated clips - Wider 3D strength and convergence controls - Full support on iPhone, iPad, Mac, and Apple Vision Pro

- max disparity fraction `0.035` (raise toward 6–8% for stronger pop); ceiling scales with eye width (`max(64, 8%)`) - strength slider honored for Toolkit Base/Large (ceiling `0.95`; recommended default `0.78`) - convergence bias slider `-0.50…0.80` (slider `0.50` = scene-centered auto median; factory default `0.25` = pop-out bias; negative / lower → more pop-out; higher → more into-screen) - stereo pop defaults migration `v15` - MV-HEVC HorizontalDisparityAdjustment remains `200` (wired via PerformanceMonitor) - short-horizon clamped 2–98 percentile normalization applied in Core ML gray remap (scene-cut reset; α≈0.12, ±2.5% slack — reduces per-frame 2–98 shake) - Spatial / Toolkit Quality temporal EMA α `0.78` with gentler motion cut (depth noise on logos no longer strips EMA) - Quality depth stride `1` (stride `2` only under memory pressure) - near-bright morphology maximum (radius 1.5); **no** Metal depth refine before warp (avoids double near-expand ghosting vs Toolkit `depthmax2` + refine stack) - StereoShift soft hole fill (mix ≤0.12) kept as quality-plus over Toolkit’s bare warp