Creating LoRas for characters can seem like a long task, but with the right tools and steps, it can be quite straightforward. In this article I'll talk about my methodology using Fancaps.net, and a custom chrome browser extension, (Which will be linked at the end or you can download it here). This method works well for low-dataset characters.
Step 1: Select Your Anime on Fancaps.net
Start by going to Fancaps.net and selecting the anime that features the character you're interested in.
Step 2: Collect Images of Your Character
Go through each episode of your selected anime and open in a new tab any images that prominently feature the character you're training the LoRa on. You can do this by middle-clicking on the images.
Step 3: Download the Images with the Extension
Once you've collected all the images you need, click on the custom browser extension and click on the "Download Images" button to download all the images you've opened in new tabs. If you don't have it download the images manually.
Step 4: Prepare Your Dataset
Follow the steps in this Civitai article to prepare your dataset. You'll be using the Dataset Maker mentioned in the article to create a dataset file and Lora trainer. Place the images you downloaded in Step 3 into this dataset directory created in the dataset maker.
Step 5: Train Your Images
Now, it's time to train your images. I aim to get the steps to be between 1500 and 2000, and always keep the epochs at 10. The formula to keep in mind is: Image Count x Number of Repeats x 10 Epochs / 2 = 1400 to 2000.
I always set the network dimension to 32 and the network alpha to 16.
Step 6: Test Your LoRa
Finally, test each LoRa from 0.5 to 0.8, at around 0.6 to 0.8 weight. Use multiple different Stable Diffusion models and test with different clothes.
And that's it! You've now created a LoRa for your character. This process can be repeated for any character from any anime. Feel free to upload images here of what you have made.
Google Chrome Extension:
(Download it here or at the link above)
(Once downloaded un-zip the file and go to extensions in Google Chrome and click on "Load Unpacked" then navigate to where the folder containing the un-zipped files.)