LoRA Under Testing, New Pics ASAP!
ps. Initial testing done, rather okay for now after rigorous testing. Will finish up later in the day
Introduction
Smolder from MLP:FiM. Actually a character I've wanted to do for a long time now, and finally got the chance to do it.
This model was trained on a large number mix of SFW and NSFW pictures, with screenshots from the show as well. Also, you're able to generate both her princess dress with makeup and her cheerleader outfit as well. See tags below.
Also, this model uses the new AI Toolkit training so results might differ from my usual. Still learning, and making sure it maintains the same quality of LoRAs I make.
Using This LoRA and want to support me? Use the 'add post' in the gallery below - This gives me a lot of blue buzz to train more LoRAs!
Illustrious (V1)
It should be able to generate:
With weightage about 0.8 to 1.0
Anthro (Put "Anthro" trigger words accordingly)
SFW and NSFW
Prompts
Trigger Words: smolder_mlp
Additional Prompts: dragon, orange scales, pale yellow underbelly, segmented underbelly, purple head spines, yellow horns, blue eyes, orange wings, tail, purple tail tip, anthro, 1girl
Cheerleader outfit: mlp_cheerleader_outfit, purple top, aqua pleated skirt, purple skirt panels, pink trim, purple pom poms
Princess outfit: princess_dress, pink lipstick, blue eyeshadow, eyeliner, eyelashes, light blue ball gown, puff sleeves, long sleeves, white trim, full skirt, pink heart tiara
Notes
Model is trained on the base Illustrious Model.
Generation quality should be about 85 - 90%+ accurate, this means you shouldn't face much issues with proper prompting.
For the princess outfit, the only issue I have is the lipstick not showing up properly with my checkpoint. If that's the case, you can check out my pictures for the LoRA i used to 'enhance' the lipstick. If not, use a checkpoint that might work better with this. (Will do more testing on this later)
For the cheerleader outfit, I seem to sometimes generate texts on the top. Include "text, letters, writing, logo, wordmark, typography" in negatives and that should solve it.
Will update more after 2nd round of testing!
Want to support me without Tipping? Here's how:
Liking my posts, collecting my images or collecting my models all generate buzz for me.
Generate more pictures of my models. I get a base commission from anyone who generates using my model, even without the tip.
Leave a like on my model. This helps me with the algorithm in reaching more people.
Leave a review. I take all feedback very seriously, good or bad.
Consider using the 'add post' under gallery... if you are posting pictures of just my LoRA. This gives me A LOT of buzz, while you can also complete your daily post quest.
Thank you very much!