This is a stylized version of Himiko Toga from My Hero Academia. This was trained from 100 custom generated images, using a custom Wai Illustrious V160 and ILXL Realism Slider (strength 3) merged checkpoint. I do not claim ownership of the original checkpoint and lora. You can now find this checkpoint on Civitai: Wai_Illustrious_ILXL_Real.
I use the following quality prompt and required tags to generate:
Positive Prompt:
ultra hd, 32K, absurdres, Masterpiece, Super Detail, 120mm lens, ULTRA REALISTIC, PERFECT EYES, PERFECT DETAILS, stylized, semi-realistic, slim body, (detailed face:1.2), large expressive eyes, soft skin,
1girl, my hero academia himiko toga, yellow eyes, slit pupils, blonde hair, double bun, messy hair, smile, teeth, fangs, brown sweater, off shoulder, tartan skirt, stockings
The Required Character Tags reproduce the trained character. Smile, teeth, fangs are optional. Clothing and expression tags are optional and can be changed freely. Because the dataset contains a variety of outfits, I recommend specifying clothing in your prompt; otherwise, the model may generate minimal or no clothing.
Negative Prompt:
text, watermark, (deformed anatomy, extra limbs, bad fingers:1.2), (disfigured, worst quality, low quality:1.2), gross proportions, (shiny skin:1.2), (deformed skin:1.2), (flat lighting:1.2), (blurry:1.2), (out of focus:1.2), (bad face details:1.2), (bad teeth:1.2), (light particles:1.2), (wet hair:1.2), (colored pupils:1.1) (large forehead:1.1), (bad hands:1.2)
I used Steps: 45, CFG: 5.5, Sampler: dpmpp_2m_sde, scheduler: karras
Training Resolution: 832 × 1344
Description
Comments (1)
This is Al Simmons here.
I’m posting a few new samples using my current flow without the V1 LoRA. I’m testing a possible Toga v2 / Chaos Gremlin retrain direction.
I wasn’t fully happy with the first run, though I thought it was close enough at the time. The goal for V2 would be stronger face consistency, yellow eyes, messy buns, petite body shape, and the unhinged grin/glare energy.
Thoughts on this direction?














