Require validated manipulation bundles

Reject filtered manipulation outputs

Expand resumable manipulation variants

Select localizer by safe region recall

Optimize localized manipulation decisions

Add locked image fusion evaluation

Optimize localizer for safe manipulation recall

Fix SD15 inpainting weight variant

Fix diffusion training dependency compatibility

Add mask-supervised manipulation localizer

Improve manipulation specialist objective

Add calibrated manipulation training fallback

Implement TruthShield high-accuracy media detector v4

Calibrate image authenticity decisions

Deploy strong AI image detector end to end

Add trained video detection pipeline and models