Experiment to explore how well diffusion models are able to capture the concept of "diffuse" subjects with diverse shapes and colors - i.e. nebulaTrained on NASA images of nebulae, some of which are "old school" and not very sharp<h3 id="lora-weights-2yv03urmk">LoRA Weights</h3>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<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/39ffbee8-aaaf-4fe8-a982-45668a85bb36/width=525/39ffbee8-aaaf-4fe8-a982-45668a85bb36.jpeg" />Three Nebulae and Dog<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bd62b64b-6611-4154-bf0f-889a97fe9d05/width=525/bd62b64b-6611-4154-bf0f-889a97fe9d05.jpeg" />

NASA Nebula

Trained on NASA images - some of which are not very sharp.

Files with SHA256: 37430b920ed3fdef0dbd31e827793b99e8d2b31b9fbccd4d16b03c96032ed77d