app.py

#!/usr/bin/env python

import os
import random
import uuid
import json
import re
import gradio as gr
import numpy as np
from PIL import Image
import spaces
import torch
from diffusers import DiffusionPipeline
from typing import Tuple
from transformers import pipeline

# Setup rules for bad words (ensure the prompts are kid-friendly)
bad_words = json.loads(os.getenv('BAD_WORDS', '["violence", "blood", "scary", "death", "ghost"]'))
default_negative = os.getenv("default_negative","")

# Add the translation pipeline
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")

def check_text(prompt, negative=""):
    for i in bad_words:
        if i in prompt:
            return True
    return False

# Kid-friendly styles
style_list = [
    {
        "name": "Cartoon",
        "prompt": "colorful cartoon {prompt}. vibrant, playful, friendly, suitable for children, highly detailed, bright colors",
        "negative_prompt": "scary, dark, violent, ugly, realistic",
    },
    {
        "name": "Children's Illustration",
        "prompt": "children's illustration {prompt}. cute, colorful, fun, simple shapes, smooth lines, highly detailed, joyful",
        "negative_prompt": "scary, dark, violent, deformed, ugly",
    },   
    {
        "name": "Sticker",
        "prompt": "children's sticker of {prompt}. bright colors, playful, high resolution, cartoonish",
        "negative_prompt": "scary, dark, violent, ugly, low resolution",
    },
    {
        "name": "Fantasy",
        "prompt": "fantasy world for children with {prompt}. magical, vibrant, friendly, beautiful, colorful",
        "negative_prompt": "dark, scary, violent, ugly, realistic",
    },
    {
        "name": "(No style)",
        "prompt": "{prompt}",
        "negative_prompt": "",
    },
]

styles = {k["name"]: (k["prompt"], k["negative_prompt"]) for k in style_list}
STYLE_NAMES = list(styles.keys())
DEFAULT_STYLE_NAME = "Sticker"

def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str, str]:
    p, n = styles.get(style_name, styles[DEFAULT_STYLE_NAME])
    return p.replace("{prompt}", positive), n + negative

DESCRIPTION = """## Children's Sticker Generator

Generate fun and playful stickers for children using AI.
"""

if not torch.cuda.is_available():
    DESCRIPTION += "\n<p>⚠️Running on CPU, This may not work on CPU.</p>"

MAX_SEED = np.iinfo(np.int32).max
CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES", "0") == "1"

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Initialize the DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained(
    "SG161222/RealVisXL_V3.0_Turbo",  # or any model of your choice
    torch_dtype=torch.float16,
    use_safetensors=True,
    variant="fp16"
).to(device)

# Convert mm to pixels for a specific DPI (300) and ensure divisible by 8
def mm_to_pixels(mm, dpi=300):
    """Convert mm to pixels and make the dimensions divisible by 8."""
    pixels = int((mm / 25.4) * dpi)
    return pixels - (pixels % 8)  # Adjust to the nearest lower multiple of 8

# Default sizes for 75mm and 35mm, rounded to nearest multiple of 8
size_map = {
    "75mm": (mm_to_pixels(75), mm_to_pixels(75)),  # 75mm in pixels at 300dpi
    "35mm": (mm_to_pixels(35), mm_to_pixels(35)),  # 35mm in pixels at 300dpi
}

# Function to post-process images (transparent or white background)
def save_image(img, background="transparent"):
    img = img.convert("RGBA")
    data = img.getdata()
    new_data = []

    if background == "transparent":
        for item in data:
            # Replace white with transparent
            if item[0] == 255 and item[1] == 255 and item[2] == 255:
                new_data.append((255, 255, 255, 0))  # Transparent
            else:
                new_data.append(item)
    elif background == "white":
        for item in data:
            new_data.append(item)  # Keep as white

    img.putdata(new_data)
    unique_name = str(uuid.uuid4()) + ".png"
    img.save(unique_name)
    return unique_name

def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    return seed

def translate_if_korean(text):
    # Check if the text contains Korean characters
    if re.search("[\uac00-\ud7a3]", text):
        # Translate Korean to English
        translation = translator(text, max_length=512)
        return translation[0]['translation_text']
    return text

@spaces.GPU(enable_queue=True)
def generate(
    prompt: str,
    negative_prompt: str = "",
    use_negative_prompt: bool = False,
    style: str = DEFAULT_STYLE_NAME,
    seed: int = 0,
    size: str = "75mm",
    guidance_scale: float = 3,
    randomize_seed: bool = False,
    background: str = "transparent",
    progress=gr.Progress(track_tqdm=True),
):
    # Translate prompt if it's in Korean
    prompt = translate_if_korean(prompt)
    
    if check_text(prompt, negative_prompt):
        raise ValueError("Prompt contains restricted words.")
    
    # Ensure prompt is 2-3 words long
    prompt = " ".join(re.findall(r'\w+', prompt)[:3])

    # Apply style
    prompt, negative_prompt = apply_style(style, prompt, negative_prompt)
    seed = int(randomize_seed_fn(seed, randomize_seed))
    generator = torch.Generator().manual_seed(seed)

    # Ensure we have only white or transparent background options
    width, height = size_map.get(size, (1024, 1024))

    if not use_negative_prompt:
        negative_prompt = ""  # type: ignore

    options = {
        "prompt": prompt,
        "negative_prompt": negative_prompt,
        "width": width,
        "height": height,
        "guidance_scale": guidance_scale,
        "num_inference_steps": 50,
        "generator": generator,
        "num_images_per_prompt": 1,  # Max 6 images
        "output_type": "pil",
    }
    
    # Generate images with the pipeline
    images = pipe(**options).images
    image_paths = [save_image(img, background) for img in images]
    
    return image_paths, seed

examples = [
    "귀여운 고양이",
    "행복한 토끼",
    "웃고있는 강아지",
    "춤추는 돌고래",
    "신나는 아기 독수리",
    "즐거운 아기 사자",    
]

css = """
footer {
    visibility: hidden;
}
"""


# Define the Gradio UI for the sticker generator
with gr.Blocks(theme="Nymbo/Nymbo_Theme", css=css) as demo:

    with gr.Group():
        with gr.Row():
            prompt = gr.Text(
                label="Enter your prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter 2-3 word prompt (e.g., cute bunny)",
                container=False,
            )
            run_button = gr.Button("Run")
        result = gr.Gallery(label="Generated Stickers", columns=2, preview=True)
    with gr.Accordion("Advanced options", open=False):
        use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=True, visible=True)
        negative_prompt = gr.Text(
            label="Negative prompt",
            max_lines=1,
            placeholder="Enter a negative prompt",
            value="(scary, violent, dark, ugly)",
            visible=True,
        )
        seed = gr.Slider(
            label="Seed",
            minimum=0,
            maximum=MAX_SEED,
            step=1,
            value=0,
            visible=True
        )
        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
        size_selection = gr.Radio(
            choices=["75mm", "35mm"],
            value="75mm",
            label="Sticker Size",
        )
        style_selection = gr.Radio(
            choices=STYLE_NAMES,
            value=DEFAULT_STYLE_NAME,
            label="Image Style",
        )
        background_selection = gr.Radio(
            choices=["transparent", "white"],
            value="transparent",
            label="Background Color",
        )
        guidance_scale = gr.Slider(
            label="Guidance Scale",
            minimum=0.1,
            maximum=20.0,
            step=0.1,
            value=6,
        )

    gr.Examples(
        examples=examples,
        inputs=prompt,
        outputs=[result, seed],
        fn=generate,
        cache_examples=CACHE_EXAMPLES,
    )

    gr.on(
        triggers=[
            prompt.submit,
            negative_prompt.submit,
            run_button.click,
        ],
        fn=generate,
        inputs=[
            prompt,
            negative_prompt,
            use_negative_prompt,
            style_selection,
            seed,
            size_selection,
            guidance_scale,
            randomize_seed,
            background_selection,
        ],
        outputs=[result, seed],
        api_name="run",
    )

if __name__ == "__main__":
    demo.queue(max_size=20).launch()