Note: I highly recommend using latent upscaling (HiResFix) + eye detailer + face detailer to fix distorted face in wider shorts. An example workflow with all fixes is attached in last image.
Main prompt tags: YuzoKoyanagiXL, 1boy, black-framed eyewear, collared shirt, necktie, suit <------ should always be included
Recommended samplers:
- Euler Ancestral Scheduler: Normal CFG: 5.0-6.5 Steps: 20 Lora strength: 1.0
Prompt tips (green - positive, red - negative):
- suit, collared shirt, necktie, dress pants - full base clothing
- (glasses:1.2) - to remove his glasses
- opaque glasses, partially opaque glasses - removes his blank glass lenses (unless you want them, then put into positives)
- 5 fingers / 4 fingers - helps controlling proper amounts of fingers
- necktie clip - canonically has no clip on his necktie, helps remove
- (handkerchief:1.3), front pocket, suit pocket - canonically none of his suits have a handkerchief in his front pocket, these prompts can help remove it but its still random
- greyscale, comic - if you want to emulate the style of the doujin
The LoRa should support all kinds of Illustrious checkpoints. Tested working well with the following:
WAI-NSFW-illustrious-SDXL v12.0
HomoSimile XL - Illustrious v2.0
To improve details:
- generate using higher base resolution (e.g. 1384x1384) or use latent 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:
Requires download of FaceDetailer custom nodes + Eyeful_v2-Paired.pt + FacesV1.pt models (can be Googled). Workflow includes a HiresFix pass with x1.30 latent upscaling by default.
For portraits/upper body shots I recommend lowering the upscaling ratio ('Upscale Latent By' node) - otherwise you might get distorted details.