Hello, it has been a long time since I created a LoRA. I have had some ideas, but little time to act on them. Today, I wanted to create for a special occasion. The game the character originates from, Oh So Hero!, has released on Steam, and reached Version 1, officially. I have enjoyed the team's efforts, and wished to create this LoRA in response. Currently, my time is still limited, so I will provide the LoRA in its current "testing" state. I plan to update the upload at a later time, with improved tagging, and attempts to avoid the style of the training data affecting the prompt. Early release, or later release, I hope you will all enjoy it. Feedback is appreciated, as well. (Versions for other models will become available when refinement is complete, but currently only NoobAI is available.)Activation keyword is joeosh; LoRA performs seemingly at weight 1, but lowering to a minimum of 0.8 is recommended if style or composition is affected heavily.Recommended positive keywords: dragon, plump, male, wings, dipstick tail, arm markings, neck markings, black eyelidsOptional keywords: fingerless gloves, mask, shorts, red thong, red sash, toplessRecommended negative keywords: panty straps, thong straps, dipstick ears(Due to the nature of the dataset, the addition of "bulge" to negatives is recommended while generating Safe-For-Work imagery.)

Joe (Oh So Hero!) LoRA

Early version trained on NoobAI, only tested on StableMondAI-SDG so far.

Files with SHA256: be6f81d1bf189c8e30677047b8479f213ba4b66aedb95ecbfe6795c02965e7b9