This Lora is used to generate cricketing shots - four types of shots can be produced - drive, pull, legglance, sweep ....
Training Images used from the dataset used - https://www.kaggle.com/datasets/aneesh10/cricket-shot-dataset
prompt :
in a cricket match, a batsman wearing <insert color> uniform plays a <insert shot type>, in a large stadium, <lora:cricket_shots:1>
negative:
baseball, deformed body, bad quality, malformed face, extra limbs
shot types are - drive, pull, legglance, sweep
it generally tends to generate cover drive, but sometimes can give straight drive also
Use baseball in negative prompt as it confuses cricket and baseball. It gives better results with 1024*1024 resolution
The faces are distorted , but using an Adetailer gives reasonable faces
I have tried using it with Juggernaut model but I guess it works with any major sdxl checkpoint.