AI

Both projects experiment with custom LoRA training pipelines in ComfyUI, using curated datasets for targeted replication of style and identity.

For the film Under Attack (left), I developed and trained a style LoRA to generate a consistent visual language.
In AI Doppelgänger (right), I combined two LoRAs — the same style model and a character LoRA trained on my own facial dataset — to explore how AI can reproduce both aesthetic and identity features within controlled generative parameters.

The final videos were produced using Kling, Higgsfield, and Veo 3, experimenting with motion, texture, and lighting continuity to extend the AI-generated visuals into cinematic movement.