This model is trained on 13k images of frame stills from various film genres, some of which had character names corrected by hand. I can't name the films here. Try it and see!
But overall this will make everything look slightly more cinema and less AI Professional Photo Shoot.
Description
Trained at a batch size of 4 with no crops. Use 1344x768 for best results. 1280x768 is weird.
Square images probably suffer and do not change toward cinematic a whole lot.
Overall, the impact is weak, which is intentional - you can scale the weights higher or use more of the trigger words to make it bigger. But, it's meant to be combined with other networks easily.