Main prompt tags: seiichikino, male focus, glasses, blue shirt <----------- should always be included
Additional tags for style choice:
for the original anime TV show style add following positive prompt: anime screenshot
to avoid the original anime style, remove anime screenshot from positive prompt or put it into negative prompt
NOTE: if you want to recreate the original style of the show without any blurriness I recommended using AutismMix SDXL checkpoint and prompting with (deblur:1.3)
Recommended sampler settings:
Sampler: Euler Ancestral Scheduler: Karras CFG: 6.5-7.5 Steps: 20-25 Lora strength: 1.0
Prompt tips:
- always include the Pony score tags in positive prompts: score_9, score_8_up, score_7_up, score_6_up etc.
- add chubby if your model doesn't generate him fat enough
- the base attire consists of: glasses, blue shirt, striped shirt (alt version)
- put shirt under shirt into negative prompt if you dont want to see a shirt under his regular shirt (this is canon tho - he wears two shirts in some episodes)
The LoRa should support all kinds of pony/SDXL checkpoints. Tested working well with the following:
AutismMix SDXL - highly recommended when going for the original TV anime style
PonyRealism
Luminaverse PonyXL
PrefectPonyXL V3
To improve details:
- generate using higher base resolution (e.g. 1216x1216) or use upscaling / highres fix - especially when trying to do full body poses
- for fixing wonky face/eyes I use FaceDetailer node (comfyui) + Adetailer Face finder model
Last image in the gallery can be imported into ComfyUI with a basic workflow:
Just Copy&Paste into ComfyUI (you might be required to download missing nodes)