I used about four hundred movie screenshots for model training, and the whole model is generally stable, if anyone else can optimize this model, allow me to thank you.

The trigger word and model name are the same