li-yan/stable-diffusion-crazy-keyboard

Are you a fan of keyboards? Above is my Cyberboard Keyboard that I used for my dialy life. It is crazy right?

Now I bring this keyboard into the Stable Diffusion, and its images can be generate by the model.

Description

This is a Stable Diffusion model fine-tuned on dataset li-yan/crazy-keyboard images.

Prompt 1: "a photo of cyberboard keyboard"

result of pretrained model	result of fine tuned model

Prompt 2: "in the forest, a cyberboard keyboard is on the ground"

result of pretrained model	result of fine tuned model

Prompt 3: "a cyberboard keyboard in front of the Golden Gate Bright"

result of pretrained model	result of fine tuned model

Details

Trained by li-yan on the li-yan/crazy-keyboard dataset.

This model is fine tuned from pretrained model CompVis/stable-diffusion-v1-4 using DreamBooth model.

More for DreamBooth model, please refer to https://dreambooth.github.io/.

Usage

pip install -qqU diffusers accelerate

import torch
from diffusers import StableDiffusionPipeline

pipe = StableDiffusionPipeline.from_pretrained("li-yan/stable-diffusion-crazy-keyboard")

# set prompt
prompt = "a photo of cyberboard keyboard" #@param
guidance_scale=12 #@param {type:"integer"}
num_inference_steps = 50 #@param {type:"integer"}

# Run the pipeline, showing some of the available arguments
pipe_output = pipe(
    prompt=prompt, # What to generate
    negative_prompt="Oversaturated, blurry, low quality", # What NOT to generate
    height=480, width=640,                                # Specify the image size
    guidance_scale=guidance_scale,                        # How strongly to follow the prompt
    num_inference_steps=num_inference_steps,              # How many steps to take
)

# View the resulting image
pipe_output.images[0]

Fine Tune Source Code