Create: LoRA Training

Custom Image model training:

With Moemate's custom image models, you can define and maintain a consistent selfie image for your character. Train a model based on some sample images and then generate any kind of images based on the trained model.

What is a LORA model

LoRA stands for Low-Rank Adaptation. It allows you to use low-rank adaptation technology to quickly fine-tune diffusion models. To put it in simple terms, the LoRA training model makes it easier to train Stable Diffusion on different concepts, such as characters or a specific style.

Training a LORA model

To train a custom image model, head to the Pro settings in the Create section of Moemate. Select “LORA Image tuning” and fill in the following fields:

LORA name (Required):

The name of your LORA can be anything you wish. Try to choose something that reminds you of the images you used to train it.

Base Model(Optional):

Choose the base model you want to use as a starting point for your LORA. Usually it's best to just keep the default option of stable-diffusion-v1.5 but you can choose any one you feel would work best. Just because you use a model as a base model when training your LORA, does not mean your LORA will only work with that model.

Fine Tuning steps(Required):

The number of iterations the AI should take when training your model. Must be a number between 100 and 5000. A good default value is around 10 steps per training image.

Learning rate(Optional):

How fast the model will be trained. 0.000001 (1e-6) is a good default. Between 1e-7 and 1e-4.

Instance prompt(Optional):

The instance prompt are the specific words you need to use when telling an image model to use this LORA. If you are training the model on an object, set this to "picture of ukj object". If you are training the model on an animal, set this to "picture of ukj animal" Add this prompt to your character's appearance, and the LORA will always be used for selfies.

Image cropping(Optional):

Training images are automatically cropped to 512x512. Center Crop will crop the center of the image. Face Crop will attempt to locate a face in the photo and crop around it. If no face is found Center Crop is used by default.

Captions(Optional): You can include a JSON file to provide a description or keywords for each of your training images. This can greatly increase the success rate of trainings

{

"myimage0001.jpeg": "ukj soda bottle on a wooden table in warm afternoon lighting, pine trees in the background",

"myimage0002.jpeg": "two ukj soda bottles next to a tropical waterfall",

}

Upload training images(Required):

Upload 5 to 200 images of the subject you want to train on. If you want accurate face results, many images of closeups of the face are needed. If you want a full body, many images of the subject's full body will be needed as well. A good number of images is around 25 to 50.

Setting up the LORA mode in Moemate:

Once you have submitted the LORA model for training, keep checking if it’s ready by refreshing the status. Once it's successful, follow these simple steps to set up the trained model for selfies.

-Refresh the Create section and head over to the “Models” section.

-Go to the “LORA” section

-Select the newly trained model.

-Go to the “Appearance” section

-Paste the prompt which you used while training the model into the appearance prompt. If you didn’t specify the instance prompt, It is “photo of ukj person” by default.

-Select any stable diffusion model from the available selfie models and try the results by clicking generate

Publish the character and enjoy the consistent selfies from the trained model.

For more Tips and tricks watch this video:

LoRA Training in UNDER 5 MINUTES in Stable Diffusion! THIS IS CRAZY!! (Create Comic Book Characters)