What version 1.2.0 improves is the , using a lightweight segmentation network to separate face from hair/neck/background more precisely.
It started on a niche subreddit dedicated to movie edits. A user named SynthDirector uploaded a clip from a classic 80s action movie. In the original, the hero gave a somber speech about war.