Image Generation Workflow

A few people have asked how I create my images, so I decided to write up my current process. This isn’t a guide, it’s just what works for me.

Steps

  1. Txt2Img or Img2Img to get the initial version of the image, using:
  1. Img2Img with ADetailer and a 2x Upscale using:
  • Checkpoint: Wai-NSFW ePred (keep the LoRAs)
  • ADetailer: person_yolov8s-seg
  • Upscaler: R-ESRGAN 4x+ Anime6B
  • Deionising Strength: 0.33 (can be adjusted)

Example Image and Prompt

The image has the metadata attached. Feel free to download it.


Prompt:
masterpiece, ultra-HD, high detail, depth of field, best quality, 8k, best quality, ergonomic, high contrast, shadows, cinematic lighting, (anime coloring, anime screencap:1.5), official art.

dusty pink hair, shag hair, messy hair, shoulder length hair, green eyes, outdoors, medium breasts, grey flannel jacket, black t-shirt, black cap, raining, hitchhiking, canvas backpack, roadside, solo, toothpick in mouth, leaning against pole,

dynamic composition, dynamic angle, <lora:lighting:1&gt, <lora:Semi-realism_illustrious:0.5>

Negative Prompt:
(lowres, worst quality, bad quality, bad anatomy, sketch), jpeg artifacts, signature, watermark, artist name, censored, bystander, logo, corner, from behind, steering wheel,

Generation Parameters
Steps: 25
Sampling Method: Euler a
Schedule Type: Automatic
CFG Scale: 5
Size: 2400x1600
Seed: 3481292005
Clip Skip: 2

Click here to see the archived version of this page, with my old processes.