This lora was trained on some horror movies. I don't really know how good it is, I think 3000 samples may not have been enough, or it might have slight overfit. For best results use h0rr0r, and then date of the movie, like 1982, 2020 or 2005, etc... then directed by James Cameron, or whoever you want, and then the prompt. You don't have to use year and director, it's just a trend I saw in the tagging and seems to give some extra flavor. Like I said, there may be overfit, so if your prompt is not very long I suggest a lower power than 1. Longer prompts seem to yield better results.

Files with SHA256: 2f05da4a54ef20b9d287ca8481508ef0501e97f2faf269ac2bd34be0c03b930e