【Stable Diffusion Mode】Models specially trained to generate realistic images
So far, we have only used the Stable Diffusion v1.5 base model to generate realistic people. Do you know there are models that are specifically trained to generate realistic images?
Things will only get better when you use them.
You will learn about some commonly used ones. We will study
Hassan blend 1.4
Realistic Vision v2
URPM I know I have missed some but bear with me.
I will use the same prompt
photo of young woman, highlight hair, sitting outside restaurant, wearing dress, rim lighting, studio lighting, looking at the camera, dslr, ultra quality, sharp focus, tack sharp, dof, film grain, Fujifilm XT3, crystal clear, 8K UHD, highly detailed glossy eyes, high detailed skin, skin pores
and the negative prompt
disfigured, ugly, bad, immature, cartoon, anime, 3d, painting, b&w
I will include the direct download link for each model. You can simply copy and paste the link to the “Model_from_URL” field in our AUTOMATIC1111 Colab notebook.
Nearly all of them are prone to generating explicit images. Use clothing terms like dress in the prompt and nude in the negative prompt to suppress them.
Some models have their own more restrictive licenses. Read them before using or incorporating them into a product.
Direct download link:https://huggingface.co/acheong08/f222/resolve/main/f222.ckpt
F222 generates realistic people with beautiful clothing. Sadly, the model is no longer in development.
Hassan blend 1.4
Direct download link:https://huggingface.co/hassanblend/hassanblend1.4/resolve/main/HassanBlend1.4_Safe.safetensors
Hassan Blend v1.4 is fine-tuned on a large number of explicit images.
Realistic Vision v2.0
Direct download link:https://civitai.com/api/download/models/29460
Realistic Vision v2 is an all-rounded model for generating photograph-style images. In addition to realistic people, it is also good for animals and scenes.
The anatomy is excellent, in my experience.
Direct download link:https://civitai.com/api/download/models/11745
Chillout Mix is the Asian counterpart of F222. It is trained to generate photo-style Asians.
Direct download link:https://huggingface.co/dreamlike-art/dreamlike-photoreal-2.0/resolve/main/dreamlike-photoreal-2.0.ckpt
Dreamlike Photoreal is an all-rounded photo-style model. The portrait images tend to be a little bit saturated.
Direct download link:https://civitai.com/api/download/models/15640
URPM is a model fine-tuned with explicit images. The anatomy is usually excellent. Images are similar to Realistic Vision v2 but a bit more polished.
In order to let you directly compare the realistic models, I used ControlNet to fix the pose. (More on this later)
The same prompt, negative prompt, and seed are used.
Which realistic model is your favorite? Let me know in the comment!
To see more content about Stable Diffusion from zero click:https://www.hayo.com/article/64c21001ef669957a0d21e63