A checkpoint has been added to the merge model to interpret natural language more accurately. As a result of this change, the method for writing prompts has been modified, making redundant quality tags unnecessary. Text within double quotation marks will be prioritized, while other words with high cosine similarity will also be automatically noted in the cardboard or sketchbook. However, a downside is that interference between tags can cause weaker tags to lose their effect due to the influence of stronger tags.