Model introduction: Using the data set of half figures to refine, the fifth training finally has a real effect;
Generation suggestion: It is recommended to use SDXL1.0 large model, the effect is undoubtedly the best; In addition, it is found that the miamodelSFWNSFWSDXL_v20 effect of c station is also very good, if downloaded to the local, can produce a good effect;
Recommended size 768x1024 or above;
Sampler method: euler a will do;
Sampling Steps: 20; Sampling steps: 20;
Prompt word guidance coefficient CFG scale: 7;
Creative introduction: realistic characters, imitation photography a scene, blue dream lights, butterflies and goldfish; From the point of view of photography, this is a close-up shot, and the face close-up will appear more clear.