app.py

# Imports
import gradio as gr
import random
import spaces
import torch
import uuid
import os

from diffusers import StableDiffusionXLPipeline, ControlNetModel
from diffusers.models import AutoencoderKL

# Pre-Initialize
DEVICE = "auto"
if DEVICE == "auto":
    DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
print(f"[SYSTEM] | Using {DEVICE} type compute device.")

# Variables
MAX_SEED = 9007199254740991
DEFAULT_INPUT = ""
DEFAULT_NEGATIVE_INPUT = "EasyNegative, deformed, distorted, disfigured, disconnected, disgusting, mutation, mutated, blur, blurry, scribble, abstract, watermark, ugly, amputation, limb, limbs, leg, legs, foot, feet, toe, toes, arm, arms, hand, hands, finger, fingers, head, heads, exposed, porn, nude, nudity, naked, nsfw"
DEFAULT_MODEL = "Default"
DEFAULT_HEIGHT = 1024
DEFAULT_WIDTH = 1024

css = '''
.gradio-container{max-width: 560px !important}
h1{text-align:center}
footer {
    visibility: hidden
}
'''

# Functions
def save_image(img, seed):
    name = f"{seed}-{uuid.uuid4()}.png"
    img.save(name)
    return name
    
def get_seed(seed):
    seed = seed.strip()
    if seed.isdigit():
        return int(seed)
    else:
        return random.randint(0, MAX_SEED)

@spaces.GPU(duration=30)
def generate(input=DEFAULT_INPUT, filter_input="", negative_input=DEFAULT_NEGATIVE_INPUT, model=DEFAULT_MODEL, height=DEFAULT_HEIGHT, width=DEFAULT_WIDTH, steps=1, guidance=0, number=1, seed=None):

    repo = None
    filter_input = filter_input || ""
    negative_input = negative_input || DEFAULT_NEGATIVE_INPUT
    seed = get_seed(seed)

    print(input, filter_input, negative_input, model, height, width, steps, guidance, number, seed)

    if model == "Anime":   
        vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
        controlnet = ControlNetModel.from_pretrained("MakiPan/controlnet-encoded-hands-130k", torch_dtype=torch.float16)
        repo = StableDiffusionXLPipeline.from_pretrained("cagliostrolab/animagine-xl-3.1", vae=vae, controlnet=controlnet, torch_dtype=torch.float16, use_safetensors=True, add_watermarker=False)
        steps = (not steps or steps < 0 and 16) or steps
        guidance = (not guidance or guidance < 0 and 7) or guidance
    else:
        vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
        controlnet = ControlNetModel.from_pretrained("MakiPan/controlnet-encoded-hands-130k", torch_dtype=torch.float16)
        repo = StableDiffusionXLPipeline.from_pretrained("sd-community/sdxl-flash", vae=vae, controlnet=controlnet, torch_dtype=torch.float16, use_safetensors=True, add_watermarker=False)
        repo.load_lora_weights("ehristoforu/dalle-3-xl-v2", adapter_name="base")
        repo.set_adapters(["base"], adapter_weights=[0.7])
        steps = (not steps or steps < 0 and 16) or steps
        guidance = (not guidance or guidance < 0 and 3) or guidance
        
    repo.to(DEVICE)
    
    parameters  = {
        "prompt": input,
        "negative_prompt": filter_input + negative_input,
        "height": height,
        "width": width,
        "num_inference_steps": steps,
        "guidance_scale": guidance,
        "num_images_per_prompt": number,
        "controlnet_conditioning_scale": 1,
        "cross_attention_kwargs": {"scale": 1},
        "generator": torch.Generator().manual_seed(seed),
        "use_resolution_binning": True,
        "output_type":"pil",
    }
    
    images = repo(**parameters).images
    image_paths = [save_image(img, seed) for img in images]
    print(image_paths)
    return image_paths

def cloud():
    print("[CLOUD] | Space maintained.")


# Initialize
with gr.Blocks(css=css) as main:
    with gr.Column():
        gr.Markdown("🪄 Generate high quality images on all styles between 10 to 20 seconds.")
        
    with gr.Column():
        input = gr.Textbox(lines=1, value=DEFAULT_INPUT, label="Input")
        filter_input = gr.Textbox(lines=1, value="", label="Input Filter")
        negative_input = gr.Textbox(lines=1, value=DEFAULT_NEGATIVE_INPUT, label="Input Negative")
        model = gr.Dropdown(label="Models", choices=["Default", "Anime"], value="Default")
        height = gr.Slider(minimum=1, maximum=2160, step=1, value=DEFAULT_HEIGHT, label="Height")
        width = gr.Slider(minimum=1, maximum=2160, step=1, value=DEFAULT_WIDTH, label="Width")
        steps = gr.Slider(minimum=-1, maximum=100, step=1, value=-1, label="Steps")
        guidance = gr.Slider(minimum=-1, maximum=100, step=0.001, value=-1, label = "Guidance")
        number = gr.Slider(minimum=1, maximum=4, step=1, value=1, label="Number")
        seed = gr.Textbox(lines=1, value="", label="Seed (Blank for random)")
        submit = gr.Button("▶")
        maintain = gr.Button("☁️")

    with gr.Column():
        images = gr.Gallery(columns=1, label="Image")
            
    submit.click(generate, inputs=[input, filter_input, negative_input, model, height, width, steps, guidance, number, seed], outputs=[images], queue=False)
    maintain.click(cloud, inputs=[], outputs=[], queue=False)

main.launch(show_api=True)