这是locon，使用插件之后调用与lora完全一样。如果你觉得难以控制，大概是你的人物lora在动作泛化上不佳。你可以调整权重比例，或者使用分层控制来进行改善。

Can you be more specific about what the problem is? Does it mean low success rate? If it can take effect, it proves that the call is successful, then you need to pay attention to the following points:

1. Model: All my training is based on NAI, please use the anime model as much as possible, and use as little as possible the model mixed with too many real models. The impact of the model is actually the biggest. If you fail repeatedly, it is better to try another model first.

2. Horizontal and vertical: try to be consistent with the illustration

3. Resolution: My training is basically 960*960, please try to generate at a nearby resolution. My general setting is 640x900

4. Multiple Loras: Try to avoid using multiple loras at the same time, especially pay attention to whether your character loras have over-fitting movements. There will also be interference between multiple loras, making your pictures more likely to collapse.
In the test, I will only add the character lora for testing, although I did not release these pictures.

5.prompt: Actually, I only noticed that you uploaded the comment picture when I wrote this. I noticed that you used a lot of negative embeddings. I actually don’t recommend doing this. The emb will change the picture to a large extent.
You can see that I didn't use emb when testing, so it may cause unexpected results.

The above are my suggestions, they are not necessarily all correct, just to provide you with ideas to try to solve