Please make sure to check the version description when using version 1.2!!!
Model Introduction: This LoRa model is inspired by the character Vladilena Milize from the anime 86 - Eighty Six. The model is trained based on the sd1.5 base model and has been tested to exhibit good versatility. However, using the same type of model for generation may yield even better results. This is the first version of the model, and the creator will continuously test, optimize, and update it. If you encounter any issues during usage, please feel free to provide feedback to the creator.
Please pay attention to the version description!!!
Recommended Weight Range: 0.7-0.9
Training Datasets and Trigger Words: This model utilizes three separate training sets, targeting the character's facial features, hat features, and clothing features. The corresponding trigger words are as follows:
Facial Features: face
Hat Features: lenahat
Clothing Features: Vladilena (note the case sensitivity)
Usage Restrictions: The use of this model and the images generated by it is strictly prohibited for any form of commercial purposes. Please comply with relevant regulations and use responsibly.
Enjoy using the model!
Description
Version 1.1 mainly addresses the issue of potentially blurry images, and during usage, I found that this model is not compatible with VAE-84000. It is recommended to use the official VAE instead.
FAQ
Comments (4)
What to do if the picture when adding Lena turns out to be in terrible quality? Promts or different models do not help
When I was experimenting with several models, I didn't have any issues with quality degradation. You can try increasing the resolution and using the parameters I provided for the image. The model used for the image in the description is Anything v3, or you can try lowering the LORA weight as well.
@Junn_ How many images is this trained on? Is it just anime images or mix with fan made stuff?
Her character is a little bit more complex, so bumping Dim to 128 and Alpha to 64 should produce better results, with some other settings adjusted.
@LittleJelly I used some animated images and found that some of the previous version's generated results were included in the training set for several versions. I will try the parameters you provided, and thank you for your message.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.