Evidence Hub

Aura: Training images, Checkpoint, and ComfyUI Workflow

Yoonjung Huh

Zenodo (CERN European Organization for Nuclear Research), 2026

Abstract

1. OverviewThis dataset supports the research article entitled “Artificial Intelligence Aura.”The study investigates the aesthetic and perceptual characteristics of images generated using a custom-trained Stable Diffusion model. The model was trained in OneTrainer using 53 original paintings created by Yoon-Jung Huh.The dataset includes the original training images and their corresponding captions, the trained model checkpoint used to generate the images examined in the study, and the ComfyUI workflow used for image generation.2. Contents01_ Training imagesThis folder contains 53 original paintings and their corresponding text captions used for model training.Each image is paired with a caption file with the same base filename.02_ CheckpointThis folder contains the trained model checkpoint used to generate the images examined in the study.The checkpoint was produced by fine-tuning Stable Diffusion XL with OneTrainer using the 53 image training dataset.03_ComfyUI_WorkflowThis folder contains the ComfyUI workflow used for image generation and captured video.3. Training InformationBase model: Stable Diffusion XL Training software: OneTrainer Number of training images: 53 Image captions: Individual text captions paired with each training image Training method: Custom checkpoint fine-tuning Training steps or epochs: [200] Image resolution: [1024*1024]5. Data ProvenanceAll 53 paintings included in the training dataset are original artworks created by Yoon-Jung Huh.The text…

Read the paper

Related studies

Go deeper in Evidence Hub

Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.

Open this study Sign up free