The image is a digital illustration of a futuristic city at night. The city is filled with tall buildings and skyscrapers, with a large planet in the background. The sky is dark and filled with stars and planets, creating a starry night sky.

In the center of the image, there is a large group of people walking through the city, some of them are carrying bags and walking towards the planet. The people are dressed in futuristic clothing, with some wearing hats and scarves. On either side of the planet, there are shops and restaurants, with large screens displaying information about the city. The overall atmosphere of the city is bustling and bustling.

breathtaking A group of people standing in a futuristic city. . award-winning, professional, highly detailed

images

samples

IMAGE

pixels

LATENT

image

model

MODEL

image_a

image_b

lora

florence2_model

clip

CONDITIONING

CLIP

text

STRING

mask

caption

data

text_positive

text_negative

CONTROL_NET

clip_l

clip_g

t5xxl

MASK

positive

negative

latent_image

control_net

git+https://github.com/WASasquatch/img2texture.git@d6159abea44a0b2cf77454d3d46962c8b21eb9d3

.mithrillion

Upscaling images using the SD3.5 blur controlnet, using Florence2 to automatically provide different text prompts for CLIP and T5 encoders. Blurring of the original image is optional and does not seem to have a large effect. Reduce the tiled diffusion batch size for lower VRAM usage. Controlnet strength seems to work best around 0.25-0.33 for slightly creative upscaling. Might have to end much earlier than 1.0 to reduce artifacts. Denoising strength might be best between 0.5-0.75 but needs more testing.

Stable Diffusion 3.5 Blur Controlnet Automatic Captioned Upscaling

OpenArt

Workflows

Active Sessions

Stable Diffusion 3.5 Blur Controlnet Automatic Captioned Upscaling

Description

Discussion

Loading...