Use this for the general "hypnotized" look. Works with anime and semi-realistic models. Works well with other LoRAs at low strengths. Don't go beyond 0.7 weight. Preview images are bad because of training image prompts, works well otherwise.

Hypnosis (General)

Hypnosis.safetensors

Use this for the general "hypnotized" look, as well as hypnosis-based situations. I tried to keep the dataset as varied as possible, so it should work for a wide range of situations. Can do both spiral and empty eyes, as well as the blank stare. Use "@_@" or "empty_eyes" tags respectively. Works well with all types of models and most LoRAs. Might require negative prompting for the best results. If you don't want the spiral background, put appropriate negative and positive tags.

Hypnosis (XL + Flux)

Trained using SD 1.5. Not as good as the XL and Flux versions, but works OK for people with performance constraints.

Files with SHA256: d864d4320fff0562e74ebd026be1e5286ea83189d997529cec58352f061c26de