FluxCraft
Train Flux LoRA
Create custom LoRAs to enhance your AI-generated images in FluxCraft
Understanding the Benefits of Style vs Regular LoRAs
A “style LoRA” is a type of LoRA (Low-Rank Adaptation) model specifically designed to capture and apply a particular artistic style, like a painting style or animation aesthetic, while a “regular LoRA” is a more general LoRA focused on adding a specific concept or feature to a model without necessarily changing the overall style drastically; essentially, a style LoRA is used to change the “look” of the generated output, while a regular LoRA focuses on adding new elements to the existing style. Key differences:- Training data:A style LoRA is trained on a dataset of images that share a consistent artistic style, like a collection of paintings by a specific artist, while a regular LoRA might be trained on images with diverse styles but a common concept (e.g., images of cats).
- Application:When applying a style LoRA, the goal is to transform the generated image to match the specific artistic style, while a regular LoRA would be used to add a particular feature or concept to the generated image without significantly changing its style.
- Style LoRA:You could use a “Van Gogh style LoRA” to make a generated image look like a painting by Vincent Van Gogh.
- Regular LoRA:You could use a “cat face LoRA” to add cat ears and whiskers to a generated image of a person, while maintaining the overall style of the base model.
Understanding the Benefits of Custom LoRAs
Training and using custom LoRAs offers numerous advantages for AI image generation:Personalized Style
Create unique artistic styles tailored to your specific vision or brand.
Specialized Subjects
Train the AI to generate specific objects, characters, or scenes with higher accuracy.
Increased Control
Gain more precise control over the output of your AI-generated images.
Efficiency
Achieve specialized results without the need for extensive model retraining.
How to Train Your Custom LoRA
Follow these steps to create your own LoRA using Pixio’s LoRA training feature.Step 1: Access LoRA Training
Step 1: Access LoRA Training
Navigate to Images -> Train Flux LoRA in the side menu at https://pixio.myapps.ai/loratraining.Note: Each LoRA training costs 400 credits.
Step 2: Set Up Your LoRA
Step 2: Set Up Your LoRA
Enter a name for your LoRA and choose a trigger word.Trigger Word: This is used to invoke your LoRA. You’ll need to include it in your prompt for the LoRA to take effect.
Step 3: Configure LoRA Type
Step 3: Configure LoRA Type
Decide whether to enable “Is Style” based on your LoRA’s purpose:
- Enable for mimicking overall aesthetic, colors, and atmosphere
- Disable for training on specific objects or characters
Step 4: Upload Training Images
Step 4: Upload Training Images
Upload at least 10 images (up to 20 for best results) that your model will train on.
Step 5: Start Training
Step 5: Start Training
Click on “Train LoRA” to begin the training process. This will take a few minutes.You’ll see the LoRAs that are currently training highlighted with a glow.
Using Your Trained LoRA
Once your LoRA is trained, you can start using it in your projects.Accessing Your LoRA
Accessing Your LoRA
Select your model from the list of Trained Models.
Utilizing Your LoRA
Utilizing Your LoRA
To use your LoRA:
- Note the trigger word
- Copy the LoRA URL
- Use both in FluxCraft to invoke the model
Tips for Effective LoRA Training
Choose Quality Images
Choose Quality Images
Select high-quality, consistent images that clearly represent the style or subject you want to train.
Vary Your Training Data
Vary Your Training Data
Include slight variations in your training images to help the LoRA generalize better.
Test Your LoRA
Test Your LoRA
After training, test your LoRA with various prompts to understand its capabilities and limitations.