Skip to content
LORA stands for Low Rank Adaptation, used to create a small dataset for LLMs.It helps in understanding and evaluating specific information to achieve desired results.
Navigate to Models and Trainings in Leonardo.Access your created datasets to view existing characters.
Ensure images are free from noise (e.g., grainy images should be removed).Keep backgrounds as plain as possible for better results.
Choose a character (e.g., Amelia) and select the Flux Dave model.Download the generated image to use for the LORA dataset.
Use ChatGPT to generate a list of images in a specific style.Provide a prompt that includes:Character nameBackground requirementsStyle specifications (e.g., cinematic, photorealistic)Aspect ratio and variety of shots.
Run the prompt in ChatGPT to receive a list of 50 shots.Approve each shot and generate images accordingly.
Go to Leonardo, select Models and Trainings, then Datasets.Choose 'Train New Model' and select 'Character'.Name the dataset and upload the generated images.
After uploading, click 'Start Training'.Ensure the base model is set to Flux Dave.
Use a trigger word (e.g., alfred_tdmb) to generate images with specific character attributes.
Assess the generated images for quality and relevance to the character.
Always check for image noise and quality before uploading.Ensure backgrounds are plain to avoid distractions in the dataset.
Use ChatGPT to quickly generate a variety of shots and expressions.Organize images in a Google Drive for easy access and management during the training process.Keep a consistent naming convention for datasets to avoid confusion.
Character LoRAs and How to Train Dataset on Leonardo AI
Everything about generating datasets and updating LoRA
Creating Character LoRAs manually
Creating a Character LORA Dataset Using Flux Dev
Objective
This SOP outlines the steps to create a character LORA dataset using the Flux Dave model, ensuring high-quality images and effective training for character consistency.
Key Steps:
1. Understanding LORA
2. Preparing Character Dataset
3. Image Quality Check
4. Creating Character Images
5. Using ChatGPT for Image Generation
6. Generating Images
7. Uploading Images to Leonardo
8. Starting Training
9. Using Trigger Words for Generation
10. Reviewing Generated Images
Note:
For Efficiency:
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ··· in the right corner or using a keyboard shortcut (
CtrlP
) instead.