Trained and tested with Flux.D BNB NF4 (12GB) in Diffusion in Low Bits set to FP16.

This is trained specifically off Philomena Cunk (Diane Morgan) in her mockumentaries, usually narrating in her Tweed jacket.

Warning: Despite my examples, this thing SUCKS at subtitle text. I mean, really, really sucks. However, the suckage could be my prompting of subtitles.

Prompting with cinematic still, high resolution photo of a woman with her hair pulled back into a ponytail with hair framing her face. She is wearing a tweed jacket