First option: Use a clean style checkpoint. If you are using a checkpoint that heavily biases its own style and it is negatively affecting the results, I would recommend using a checkpoint that is more neutral in terms of style. You can review the provided examples and choose a checkpoint that better fits your needs. [Leave examples in the comments: A good example in my case turned out to be "Anything"]
Second option: Load additional relevant information into the program. If you are inputting information that the model did not train on in Lora, you might be introducing too much influential information on the style. Try using prompts that are more relevant to the task you are performing and avoid introducing additional information that may affect the desired style.
Third option: Increase the intensity in Lora. If you are getting unsatisfactory results, you can try increasing the intensity in Lora. You can use a value greater than 1, like Lora:1.2, to obtain more prominent results in terms of style. Experiment with different values to find the right balance.
Fourth option: Avoid using a VAE. If you are using a Variational Autoencoder (VAE), keep in mind that this type of model can affect the colorimetry of the image and break the style stored in Lora. Consider not using a VAE if you want to preserve the original style more accurately.
Fifth option: Gather more information and feedback to get a more comprehensive opinion. If you are experiencing difficulties in achieving the desired styles in LORA, an additional option is to seek more information and opinions from other users who have encountered similar issues. Gathering different perspectives can provide valuable insights and alternative solutions to improve the situation. Do not underestimate the power of the community in problem-solving and generating innovative ideas. [Leave your contribution in the comments, create this fifth option by providing your homemade solutions]
Financial assistance: Hello everyone!
This is Tomas Agilar speaking, and I'm thrilled to have the opportunity to share my work and passion with all of you. If you enjoy what I do and would like to support me, there are a few ways you can do so: