Trained on a large dataset of 500+ images of Gideon aka Mark Parsons
Recommended base model: Indugo Furry mix v45_hybrid (and similar furry models)
Also available to quickly try on SeaArt: https://www.seaart.ai/models/detail/128a8e64abf88e92aa2c5dd9fe760d65
v1.0: Small layer count model, not the most stable but captures the gideon flavour nicely
v2.0: Larger 128 layer model. More stable but different results, so feel free to try both
v3.2: New learn setting on better hardware. Works MUCH better.
Generic model that should handle anything, it is vaguely familiar with some of his characters by name. Backgrounds will be mostly simple since that is what the artist typically draws. Human characters work do not look gideon like (no training data for them). Since the dataset topic is broad you do want to be descriptive in the prompt (species, body type, colors, clothes, scene..etc) to avoid instability due to mixing too many ideas. Tags used are e621 and boru.
Models influence is fairly strong so use it with a weight of around 0.6 to 0.9. You can use the tags 'simple shading' or 'flat shading' to get a less shaded look in models that are too biased towards realistic shading.
Keywords: gideon, gideon style
Description
Larger 128 layer model trained on the same 500+ image dataset
More stable than v1.0 but has a different feel, so give both a try.
FAQ
Comments (3)
I'm getting bad results. Any tips please?
- Use a good furry base model (It does NOT work well in a generic StableDifusion model)
- Reduce the weight of this LoRA (default weight of 1.00 is too much for this one)
- Mention gideon and/or gideon style in the prompt
- Try adding simple shading or flat colors to the prompt (his art style is not very detailed, but some models try to do detailed by default)
- Be descriptive in the prompt (like body shape, clothes or describe what they look like naked..etc)
- Try with species that Gideon offten draws first.
@berni__ I think I got it, Thanks.