Higher quality 2.1 images!
Thank you for checking out the new and improved digital diffusion! This model is a general purpose 2.1 model that does well in many styles and themes. I really hope you enjoy the model, and please share what you can create with it!
Basic prompting guide:
This model might prompt closer to SD 1.5, since you don't need overly complicated prompts to get really cool images. The two most important tokens are digital art and photograph. These control what kind of image you are generating. It's pretty self explanatory what they do as well.
There are other tokens that can improve your images or make them feel a certain way, or are just fun to experiment with. Here are some of my favorites!
Portrait
Landcsape
Vehicle
Cinematic
Dramatic
Fantasy
Particles
Steampunk
Moody
Dark
Bright
Calm
Sunrise
FOR V1: the prompting guide to better understand how to use the model to it's fullest extent, or at least to the extent I have discovered.
Description
IMPROVED PHOTOREALISM -
I overhauled the dataset by removing 100 images, and adding 150 more, bringing the total to 300 images. I captioned all photography with the tag "photograph" and all digital art with the tag "digital art". Now can do professional like photography as well as digital art, both very well!
Trained on top of v1.
FAQ
Comments (7)
I don't know why but all my outputs are black :(
Weird, do you have the config file set up?
If you are using AUTOMATIC1111 webui, go to settings -> Stable diffusion -> enable "Upcast cross attention layer to float32" , this should be enabled when using 2.1 models. Disable it for 1.5 models.
Great model and great results, kudos to the creator
Thank you! Means a lot.
Does this do NSFW things well?
Not to my knowledge, I didn't train it on that. I suggest looking for models specifically trained on nsfw content if you are looking for that sort of thing in 2.1
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.