AETERNITAS utilizes SUPIR v2 (Tiled) to "hallucinate" high-frequency textures back into existence. This overcomes the "waxy" artifacts common in quantized models, restoring pores, fabric weave, and realistic lighting at 4K resolution.
Using the Sonic v2 animation driver, the system anticipates speech patterns and drives micro-expressions—such as squinting before a shout or muscle tension before a sigh—creating a "soulful" digital presence that transcends traditional lip-sync.
Optimized for consumer hardware, the pipeline leverages a Wan 2.2 14B MoE architecture quantized to Q3_K_S, achieving professional-grade volumetric output on a single RTX 3060 Ti budget.