-
Fix active-speaker face/lip: per-frame speaker selection; drop match_seq==2; scalars = frame-0 aliases
-
Add derived training artifacts: alignnorm npz, per-frame lip sidecar, keep-list (8,884) (#3)
Haosonnn +1: hao3on
-
Switch annotations from MonST3R to MegaSaM 001-030 (metric, w2c, per-frame speaker face/lip)
-
Add monst3r pose + face/lip-bbox annotations (parts 001-040)
