SemiRealMix

The result of many merges aimed at making semi-realistic human images.

I use the following options to get good generation results:

Prompt:

delicate, masterpiece, best shadow, (1 girl:1.3), (korean girl:1.2), (from side:1.2), (from below:0.5), (photorealistic:1.5), extremely detailed skin, studio, beige background, warm soft light, low contrast, head tilt

Negative prompt:

(worst quality, low quality:1.4), nsfw, nude, (loli, child, infant, baby:1.5), jewely, (hard light:1.5), back light, spot light, hight contrast, (eyelid:1.3), outdoor, monochrome

Sampler: DPM++ SDE Karras

CFG Scale: 7

Steps: 20

Size: 512x768

Denoising strength: 0.5, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Eta: 0.2

Clip skip: 2

Base Model : SD 1.5

VAE: vae-ft-mse-840000-ema-pruned

Use xformers : True

🧨 Diffusers

This model can be used just like any other Stable Diffusion model. For more information, please have a look at the Stable Diffusion.

You can also export the model to ONNX, MPS and/or FLAX/JAX.

from diffusers import StableDiffusionPipeline
import torch

model_id = "robotjung/SemiRealMix"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "1girl"
image = pipe(prompt).images[0]

image.save("./output.png")

Examples:

Here are some examples of images generated using this model: