B/W Photography Creator: 1967371825646

This model was trained by me using lokr, with around 10,000 images as the dataset. Later, I integrated a model from a caracal creator to reduce limb distortion. The results are quite satisfactory.

Since most of the dataset consists of human figures, performance in other aspects is weaker. Sampling methods: DMP++2M Karras or 3m algorithm.

Iterations: 30 to 50.

Recommended dimensions: 1:1 / 2:3 / 3:2 / 3:4 / 4:3 / 9:16 / 16:9 (other dimensions are welcome to try).

Trigger word: black-and-white photograph

Prompt guidance coefficient (CFG): 7.0.