app.py

import gradio as gr
import requests
import io
import random
import os
from PIL import Image

# List of available models
list_models = [
    "SDXL 1.0", "SD 1.5", "OpenJourney", "Anything V4.0",
    "Disney Pixar Cartoon", "Pixel Art XL", "Dalle 3 XL",
    "Midjourney V4 XL", "Open Diffusion V1", "SSD 1B",
    "Segmind Vega", "Animagine XL-2.0", "Animagine XL-3.0",
    "OpenDalle", "OpenDalle V1.1", "PlaygroundV2 1024px aesthetic",
]

# Function to generate images from text
def generate_txt2img(current_model, prompt, is_negative=False, image_style="None style", steps=50, cfg_scale=7, seed=None):

    if current_model == "SD 1.5":
        API_URL = "https://api-inference.huggingface.co/models/runwayml/stable-diffusion-v1-5"
    elif current_model == "SDXL 1.0":
        API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-xl-base-1.0"
    elif current_model == "OpenJourney":
        API_URL = "https://api-inference.huggingface.co/models/prompthero/openjourney"       
    elif current_model == "Anything V4.0":
        API_URL = "https://api-inference.huggingface.co/models/xyn-ai/anything-v4.0" 
    elif current_model == "Disney Pixar Cartoon":
        API_URL = "https://api-inference.huggingface.co/models/stablediffusionapi/disney-pixar-cartoon"
    elif current_model == "Pixel Art XL":
        API_URL = "https://api-inference.huggingface.co/models/nerijs/pixel-art-xl"
    elif current_model == "Dalle 3 XL":
        API_URL = "https://api-inference.huggingface.co/models/openskyml/dalle-3-xl"
    elif current_model == "Midjourney V4 XL":
        API_URL = "https://api-inference.huggingface.co/models/openskyml/midjourney-v4-xl"
    elif current_model == "Open Diffusion V1":
        API_URL = "https://api-inference.huggingface.co/models/openskyml/open-diffusion-v1" 
    elif current_model == "SSD 1B":
        API_URL = "https://api-inference.huggingface.co/models/segmind/SSD-1B"
    elif current_model == "Segmind Vega":
        API_URL = "https://api-inference.huggingface.co/models/segmind/Segmind-Vega"
    elif current_model == "Animagine XL-2.0":
        API_URL = "https://api-inference.huggingface.co/models/Linaqruf/animagine-xl-2.0"
    elif current_model == "Animagine XL-3.0":
        API_URL = "https://api-inference.huggingface.co/models/cagliostrolab/animagine-xl-3.0"    
    elif current_model == "OpenDalle":
        API_URL = "https://api-inference.huggingface.co/models/dataautogpt3/OpenDalle"
    elif current_model == "OpenDalle V1.1":
        API_URL = "https://api-inference.huggingface.co/models/dataautogpt3/OpenDalleV1.1" 
    elif current_model == "PlaygroundV2 1024px aesthetic":
        API_URL = "https://api-inference.huggingface.co/models/playgroundai/playground-v2-1024px-aesthetic"  

        
    API_TOKEN = os.environ.get("HF_READ_TOKEN")
    headers = {"Authorization": f"Bearer {API_TOKEN}"}


    if image_style == "None style":
        payload = {
            "inputs": prompt + ", 8k",
            "is_negative": is_negative,
            "steps": steps,
            "cfg_scale": cfg_scale,
            "seed": seed if seed is not None else random.randint(-1, 2147483647)
        }
    elif image_style == "Cinematic":
        payload = {
            "inputs": prompt + ", realistic, detailed, textured, skin, hair, eyes, by Alex Huguet, Mike Hill, Ian Spriggs, JaeCheol Park, Marek Denko",
            "is_negative": is_negative + ", abstract, cartoon, stylized",
            "steps": steps,
            "cfg_scale": cfg_scale,
            "seed": seed if seed is not None else random.randint(-1, 2147483647)
        }
    elif image_style == "Digital Art":
        payload = {
            "inputs": prompt + ", faded , vintage , nostalgic , by Jose Villa , Elizabeth Messina , Ryan Brenizer , Jonas Peterson , Jasmine Star",
            "is_negative": is_negative + ", sharp , modern , bright",
            "steps": steps,
            "cfg_scale": cfg_scale,
            "seed": seed if seed is not None else random.randint(-1, 2147483647)
        }
    elif image_style == "Portrait":
        payload = {
            "inputs": prompt + ", soft light, sharp, exposure blend, medium shot, bokeh, (hdr:1.4), high contrast, (cinematic, teal and orange:0.85), (muted colors, dim colors, soothing tones:1.3), low saturation, (hyperdetailed:1.2), (noir:0.4), (natural skin texture, hyperrealism, soft light, sharp:1.2)",
            "is_negative": is_negative,
            "steps": steps,
            "cfg_scale": cfg_scale,
            "seed": seed if seed is not None else random.randint(-1, 2147483647)
        }

    image_bytes = requests.post(API_URL, headers=headers, json=payload).content
    image = Image.open(io.BytesIO(image_bytes))
    return image

# Function to read CSS from file
def read_css_from_file(filename):
    with open(filename, "r") as file:
        return file.read()

# Read CSS from file
css = read_css_from_file("style.css")

PTI_SD_DESCRIPTION = '''
<div id="content_align">
  <span style="color:darkred;font-size:32px;font-weight:bold">  
    Stable Diffusion Models Image Generation
  </span>
</div>
<div id="content_align">
  <span style="color:blue;font-size:16px;font-weight:bold">  
    Generate images directly from text prompts (no parameter tuning required)
  </span>
</div>
<div id="content_align" style="margin-top: 10px;">
</div>
'''
# Prompt examples
#prompt_examples = [
#    "A blue jay standing on a large basket of rainbow macarons.",
#    "A dog looking curiously in the mirror, seeing a cat.",
#    "A robot couple fine dining with Eiffel Tower in the background.",
#    "A chrome-plated duck with a golden beak arguing with an angry turtle in a forest.",
#    "A transparent sculpture of a duck made out of glass. The sculpture is in front of a painting of a landscape.",
#    "A cute corgi lives in a house made out of sushi.",
#    "A single beam of light enter the room from the ceiling. The beam of light is illuminating an easel. On the easel there is a Rembrandt painting of a raccoon.",
#    "A photo of a Corgi dog riding a bike in Times Square. It is wearing sunglasses and a beach hat."]



# Creating Gradio interface
with gr.Blocks(css=css) as demo:
    gr.Markdown(PTI_SD_DESCRIPTION)
    with gr.Row():   
        with gr.Column():
            current_model = gr.Dropdown(label="Select Model", choices=list_models, value=list_models[1])
            text_prompt = gr.Textbox(label="Input Prompt", placeholder="Example: A blue jay ", lines=2)
        with gr.Column():
            negative_prompt = gr.Textbox(label="Negative Prompt (optional)", placeholder="Example: blurry, unfocused", lines=2)
            image_style = gr.Dropdown(label="Select Style", choices=["None style", "Cinematic", "Digital Art", "Portrait"], value="None style")
                
        generate_button = gr.Button("Generate Image", variant='primary')
        
    with gr.Row():
        image_output = gr.Image(type="pil", label="Image Output")

    generate_button.click(generate_txt2img, inputs=[current_model, text_prompt, negative_prompt, image_style], outputs=image_output)

# Launch the app
demo.launch()