|
import gradio as gr |
|
from fetch import get_values |
|
from dotenv import load_dotenv |
|
load_dotenv() |
|
import prodia |
|
import requests |
|
import random |
|
from datetime import datetime |
|
import os |
|
|
|
|
|
|
|
prodia_key = os.getenv('PRODIA_X_KEY', None) |
|
if prodia_key is None: |
|
print("Please set PRODIA_X_KEY in .env, closing...") |
|
exit() |
|
client = prodia.Client(api_key=prodia_key) |
|
|
|
def process_input_text2img(prompt, negative_prompt, steps, cfg_scale, number, seed, model, sampler, aspect_ratio, upscale, save=False): |
|
images = [] |
|
for image in range(number): |
|
result = client.txt2img(prompt=prompt, negative_prompt=negative_prompt, model=model, sampler=sampler, |
|
steps=steps, cfg_scale=cfg_scale, seed=seed, aspect_ratio=aspect_ratio, upscale=upscale) |
|
images.append(result.url) |
|
if save: |
|
date = datetime.now() |
|
if not os.path.isdir(f'./outputs/{date.year}-{date.month}-{date.day}'): |
|
os.mkdir(f'./outputs/{date.year}-{date.month}-{date.day}') |
|
img_data = requests.get(result.url).content |
|
with open(f"./outputs/{date.year}-{date.month}-{date.day}/{random.randint(1, 10000000000000)}_{result.seed}.png", "wb") as f: |
|
f.write(img_data) |
|
return images |
|
|
|
def process_input_img2img(init, prompt, negative_prompt, steps, cfg_scale, number, seed, model, sampler, ds, upscale, save): |
|
images = [] |
|
for image in range(number): |
|
result = client.img2img(imageUrl=init, prompt=prompt, negative_prompt=negative_prompt, model=model, sampler=sampler, |
|
steps=steps, cfg_scale=cfg_scale, seed=seed, denoising_strength=ds, upscale=upscale) |
|
images.append(result.url) |
|
if save: |
|
date = datetime.now() |
|
if not os.path.isdir(f'./outputs/{date.year}-{date.month}-{date.day}'): |
|
os.mkdir(f'./outputs/{date.year}-{date.month}-{date.day}') |
|
img_data = requests.get(result.url).content |
|
with open(f"./outputs/{date.year}-{date.month}-{date.day}/{random.randint(1, 10000000000000)}_{result.seed}.png", "wb") as f: |
|
f.write(img_data) |
|
return images |
|
|
|
""" |
|
def process_input_control(init, prompt, negative_prompt, steps, cfg_scale, number, seed, model, control_model, sampler): |
|
images = [] |
|
for image in range(number): |
|
result = client.controlnet(imageUrl=init, prompt=prompt, negative_prompt=negative_prompt, model=model, sampler=sampler, |
|
steps=steps, cfg_scale=cfg_scale, seed=seed, controlnet_model=control_model) |
|
images.append(result.url) |
|
return images |
|
""" |
|
|
|
theme = "Base" |
|
|
|
|
|
with gr.Blocks(theme=theme) as demo: |
|
gr.Markdown(""" |
|
# Stable Diffusion Demo |
|
|
|
<h3></h3> |
|
|
|
|
|
🚀 This space generates images by text with many settings! |
|
|
|
⏰️ Generation on average lasts 15-25 seconds! |
|
|
|
👥️️ This demo was created by OpenskyML and 4COM! |
|
|
|
|
|
|
|
""") |
|
|
|
gr.Image("banner.png", elem_id="banner-image", show_label=False, show_download_button=False, show_share_button=False) |
|
|
|
|
|
|
|
gr.DuplicateButton(value="Duplicate space for private use") |
|
|
|
|
|
|
|
with gr.Tab(label="txt2img"): |
|
with gr.Row(): |
|
with gr.Column(): |
|
prompt = gr.Textbox(label="Prompt", lines=2, placeholder="beautiful cat, 8k") |
|
negative = gr.Textbox(label="Negative Prompt", lines=3, value="text, blurry, fuzziness", placeholder="Add words you don't want to show up in your art...") |
|
|
|
with gr.Row(): |
|
steps = gr.Slider(label="Steps", value=30, step=1, maximum=50, minimum=5, interactive=True) |
|
cfg = gr.Slider(label="CFG Scale", maximum=20, minimum=1, value=7, interactive=True, info="Recommended 7 CFG Scale") |
|
|
|
with gr.Row(): |
|
num = gr.Slider(label="Number of images", value=2, step=1, maximum=4, minimum=1, interactive=True) |
|
seed = gr.Slider(label="Seed", value=-1, step=1, minimum=-1, maximum=4294967295, interactive=True, info="""'-1' is a random seed""") |
|
|
|
with gr.Row(): |
|
model = gr.Dropdown(label="Model", choices=get_values()[0], value="v1-5-pruned-emaonly.ckpt [81761151]", interactive=True) |
|
sampler = gr.Dropdown(label="Sampler", choices=get_values()[1], value="DPM++ SDE Karras", interactive=True) |
|
|
|
with gr.Row(): |
|
ar = gr.Radio(label="Aspect Ratio", choices=["square", "portrait", "landscape"], value="square", interactive=True) |
|
with gr.Column(): |
|
upscale = gr.Checkbox(label="upscale", value=True, interactive=True, info="""'True' recommended, improves image quality""") |
|
|
|
with gr.Row(): |
|
run_btn = gr.Button("Generate", variant="primary") |
|
with gr.Column(): |
|
result_image = gr.Gallery(label="Result Image(s)") |
|
|
|
gr.Examples( |
|
examples=[ |
|
["A high tech solarpunk utopia in the Amazon rainforest"], |
|
["A pikachu fine dining with a view to the Eiffel Tower"], |
|
["A mecha robot in a favela in expressionist style"], |
|
["an insect robot preparing a delicious meal"], |
|
["A small cabin on top of a snowy mountain in the style of Disney, artstation"] |
|
], |
|
|
|
inputs=[prompt], |
|
cache_examples=False, |
|
) |
|
|
|
run_btn.click( |
|
process_input_text2img, |
|
inputs=[ |
|
prompt, |
|
negative, |
|
steps, |
|
cfg, |
|
num, |
|
seed, |
|
model, |
|
sampler, |
|
ar, |
|
upscale |
|
], |
|
outputs=[result_image], |
|
) |
|
|
|
with gr.Tab(label="img2img"): |
|
with gr.Row(): |
|
with gr.Column(): |
|
prompt = gr.Textbox(label="Prompt", lines=2, placeholder="beautiful cat, 8k") |
|
|
|
with gr.Row(): |
|
negative = gr.Textbox(label="Negative Prompt", lines=3, placeholder="Add words you don't want to show up in your art...") |
|
init_image = gr.Textbox(label="Init Image Url", lines=3, placeholder="https://cdn.openai.com/API/images/guides/image_generation_simple.webp") |
|
|
|
|
|
with gr.Row(): |
|
steps = gr.Slider(label="Steps", value=30, step=1, maximum=50, minimum=1, interactive=True) |
|
cfg = gr.Slider(label="CFG Scale", maximum=20, minimum=1, value=7, interactive=True, info="Recommended 7 CFG Scale") |
|
|
|
with gr.Row(): |
|
num = gr.Slider(label="Number of images", value=2, step=1, maximum=4, minimum=1, interactive=True) |
|
seed = gr.Slider(label="Seed", value=-1, step=1, minimum=-1, maximum=4294967295, interactive=True, info="""'-1' is a random seed""") |
|
|
|
with gr.Row(): |
|
model = gr.Dropdown(label="Model", choices=get_values()[0], value="v1-5-pruned-emaonly.ckpt [81761151]", interactive=True) |
|
sampler = gr.Dropdown(label="Sampler", choices=get_values()[1], value="DPM++ 2M Karras", interactive=True) |
|
|
|
with gr.Row(): |
|
ds = gr.Slider(label="Denoising strength", maximum=0.9, minimum=0.1, value=0.5, interactive=True) |
|
with gr.Column(): |
|
upscale = gr.Checkbox(label="upscale", value=True, interactive=True, info="""'True' recommended, improves image quality""") |
|
|
|
|
|
with gr.Row(): |
|
run_btn = gr.Button("Generate", variant="primary") |
|
with gr.Column(): |
|
result_image = gr.Gallery(label="Result Image(s)") |
|
run_btn.click( |
|
process_input_img2img, |
|
inputs=[ |
|
init_image, |
|
prompt, |
|
negative, |
|
steps, |
|
cfg, |
|
num, |
|
seed, |
|
model, |
|
sampler, |
|
ds, |
|
upscale |
|
], |
|
outputs=[result_image], |
|
) |
|
|
|
with gr.Tab(label="Gallery"): |
|
|
|
gr.load("nateraw/stable_diffusion_gallery", src="spaces") |
|
|
|
with gr.Tab(label="License"): |
|
|
|
gr.load("4com/4com-license", src="spaces") |
|
|
|
if __name__ == "__main__": |
|
demo.launch(show_api=True) |