Generate a 60-Image LoRA Dataset from a Single Character Image

This workflow creates a rich, varied dataset of 60 images from a single input image. It includes a range of poses, lighting, expressions, and environments to give your LoRA maximum flexibility and generalization.

Key Inputs

Load Image: Upload your character image here. Best results come from character sheets with multiple angles (front and back), but a front-facing image will also work well.

Character Name (Output Folder): Set the name of the character — this also becomes the name of the output folder where your generated images will be saved. This helps keep things organized for LoRA training.

Prompt List: This controls the variety of images generated. The default list includes 60 diverse prompts from ChatGPT, but you can freely edit it to add or remove prompts. Want a specific setting or theme? Just replace or expand the prompt list with your own.

Not every image will be perfect, but the ideal minimum for a good Flux character LoRA dataset is about 30 images, so out of the 60 images pick only the best ones to train your LoRA!