sky bar:The 1.0 version of the model is easy to produce close shots, while the 1.01 version is easy to produce mid range shots. Same training data for both versions, so I uploaded both I personally feel that V1.0 has a better fusion of various prompt wordweights ranging from 0.5 to 1.0. Training use 768×512. The dataset is all images of unmanned scenes in bar, so the effect of a separate unmanned scene image is better. If you want to add man, you can reduce the weight. Adding characters such as Lora or adding characters/nightclubs/music festivals to the prompt words can also produce some special effects, Remember to use Adetailer to repair the face. The online image quality is not 4k, and if some images are unclear, high-definition repair can be added.1.0 1.01 V1.0 0.5-1.0 768×512 Lora / / Adetailer 4k