| Recommended Usage & Tips | |
| All Dataset Focus: The training dataset is primarily focused on facial features. | |
| Best Settings: To get the best results, use a LoRA weight of 0.6 to 0.8. | |
| No Trigger Word Needed: You do not need to include any trigger words in your prompt. (Even though the trigger word defaults to the file name, it is best not to use it). | |
| 💡 Why? | |
| Since the dataset is heavily focused on the face, using the trigger word or a high weight can make the image look forced or unnatural. Leaving out the trigger word | |
| and keeping the weight between 0.6 and 0.8 will allow the LoRA to blend seamlessly with the base model, creating a much more natural and high-quality image. |