Experiment to explore how well diffusion models are able to capture the concept of "diffuse" subjects with diverse shapes and colors - i.e. nebula
Trained on NASA images of nebulae, some of which are "old school" and not very sharp
LoRA Weights
As expected of the nebulous edges of the subject, nebula composition changes with LoRA weight. Higher weights will increasingly converge the nebula to show nebula structure and/or sources.
Single Nebula and Boy

Three Nebulae and Dog

Description
Trained on NASA images - some of which are not very sharp.
Details
Downloads
136
Platform
CivitAI
Platform Status
Available
Created
9/13/2024
Updated
11/12/2025
Deleted
-
Trigger Words:
nebula
Files
837857_training_data.zip
Mirrors
CivitAI (1 mirrors)
Nebula.safetensors
Mirrors
CivitAI (1 mirrors)