BLOG

Roman Baths 3

by GymDreams
Stable Diffusion with SDXL
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.
Roman Baths 3. Oil painting. Stable Diffusion with SDXL.

Originally, this one was in the reject pile, but I decided to see if I could save it by putting the first render into another img2img pass with a 0.3 denoise, and further applied ADetailer to all 9 faces. That took a very long time, but the results are a lot better now, and I think that most likely I will include this in the final series.

That said, if you zoom into the image, then you’ll see that there are clearly issues with the men in the background. I will likely inpaint all of them in Stable Diffusion in the future. It will be a very tedious process, but I think that it will be worth it.

  1. Stable Diffusion img2img from MJ render with 0.9 Denoise, 50 steps, DPM++ 2M SDE Karras, SDXL Base, SDXL VAE
  2. Stable Diffusion img2img from SDXL render with 0.3 Denoise, 50 steps, DPM++ 3M SDE Karras, SDXL Base, SDXL VAE
  • Control Net dw_openpose_full, thibaud_xl_openpose_256lora, Start 0, End 0.5, Balanced
  • Automatic1111 v1.6.0

Roman Baths 3
Oil painting
Stable Diffusion with SDXL

  • Rendered 1: 2023-10-21
  • Rendered 2: 2023-11-04
  • Published: 2023-11-04

Tech

Steps: 50, Sampler: DPM++ 3M SDE Exponential, CFG scale: 6, Seed: 3254044941, Size: 1536x1536, Model hash: 31e35c80fc, Model: sd_xl_base_1.0, VAE hash: 63aeecb90f, VAE: sdxl_vae.safetensors, Denoising strength: 0.9, ControlNet 0: "Module: dw_openpose_full, Model: thibaud_xl_openpose_256lora (14288071), Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 0.5, Pixel Perfect: False, Control Mode: Balanced", Version: v1.6.0