luna

Stable diffusion is a deep learning, text-to-image model and used to generate detailted images conditioned on text description, thout it can also be applied to other task such as inpainting or outpainting and generate image to image translate guide by text prompt.

try online on google colab:

luna can running on online colab notebook you can check here:

• luna
• luna with tpu
• luna with gpu and mixed precision
  • 15 seconds generation time per image (512 x 512) on default colab GPU without drop the quality

usagagge for online colab:

• click the connect

  

• open on runtime and click run all ( ctrl+f9 if using shortcut )

  

• online collab running luna

usage

use venv

//create venv
python3 -m venv venv

//activate venv
source venv/bin/activate

clone repo

clone on https :

git clone https://github.com/De-Rune/luna
cd luna
pip install -r requirements.txt

clone on ssh :

git clone [email protected]:De-Rune/luna.git
cd luna
pip install -r requirements.txt

note : if using mac m1 you can try installing the requirements_m1.txt

run script

python text2image.py --prompt="example text"

to change the output file name run using the --output flag

python text2image.py --prompt="cool picture" --output="cool_pic.png"

install as python package

pip install git+https://github.com/De-Rune/luna

and run the package using

from stable_diffusion_tensorflow.stable_diffusion import StableDiffusion

generator = StableDiffusion(img_height=512, img_width=512, jit_compose=False)
img = generator.generate(
  "DSLR photograph of an astronut riding a horse",
  num_steps = 50,
  unconditional_guidance_scale = 75,
  temperature = 1,
  batch_size = 1,
)

you can change dimension of image by change the img_height and img_width

generator = StableDiffusion(
  img_height = 1020 # or change 1080
  img_height = 1080 # or change 800
)

⚠️ NOTE ⚠️

if pip encountered an issue, try running pip with higher privilege using sudo

examples

prompt	image
minimalist house with family, mountainous forested wild, concept art illustration
natural cave wall, dynamic light, mist low over ground, illustration by josan gonzales and moebius, studio muti, malika favre, rhads, makoto, clean thick line, comics style
A beautiful ultradetailed anime illustration of a city street, trending on artstation

---