FaceForgeToolset


Set	FaceForge
Role	toolset
Latest	`0.2.2` — 11 September 2026
Versions here	5
Depends on	`FaceForge`
Source	github.com/AutomationForgeHQ/FaceForgeToolset

0.2.x

CloneFaceBankForLanguage records the source face bank on the clone. That is what lets the clone's clips follow the source-language performance underneath a dub instead of guessing the acting from audio.
The version this toolset reports to an agent is now read from the plugin's own descriptor rather than repeated in C++, so it can no longer answer a number the installed package does not carry.

Packaging fix: the release now carries everything the register allows. BuildPlugin's filter excludes Config/ and every public_extra path, so earlier zips shipped without them.
GetToolsetVersion() answers this plugin's real version; it had drifted from the descriptor.

Face clips gain layers; the mix is a recipe, not a rerun
Merge crosses vocabularies; mouth layers override only what they drive
The culmination wiring: one library, a teleprompter, and panels that meet
Casting made human-readable and controllable: audition, discovered actions, a stage
A re-voicing given its own path, distinct from a new take
The take became the unit: four quadrants, folders, one decision
Hear a voice without committing to it, and a face that gets made
Face bank now links to its speech bank by id instead of guessing
Speech and face joined by id: a speaker names their head, a bank says what it holds
Hand correction shipped as a layer, so the solve underneath can be replaced
Localisation: a sibling bank per language, joined by line id
A face bank now answers what a line plays, and takes outside buttons
Repointed to kovati.dev

Apache-2.0 relicensing
Every plugin descriptor made to agree with its release tag
FaceForgeACE shipped: a MetaHuman can now be baked, and the container gained controls
Versions cut for the release