Gemma 3 encoder (also called Gemma 3 12B, or Text encoder) is a term in Cinematography, a type of model architecture. The instruction-tuned text encoder (3,840 hidden dims, 262,208 vocab) parsing prompts into the shared latent manifold.
Numonic normalises cinematography vocabulary like this so shots, camera moves, and lighting stay searchable and reproducible across AI-generated footage.
Related Terms
See AI Asset Management in Action
Numonic automatically captures provenance, preserves metadata, and makes every AI-generated asset searchable and reproducible.