app.py

import gradio as gr
import os
from PIL import Image, ImageDraw
import re
from io import BytesIO

from huggingface_hub import InferenceClient
from diffusers import StableDiffusionPipeline
import torch

client = InferenceClient()

pipe = StableDiffusionPipeline.from_pretrained(
    "CompVis/stable-diffusion-v1-4",
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
)
pipe = pipe.to("cuda" if torch.cuda.is_available() else "cpu")


def screenwriter(prompt: str) -> str:
    instructions = f"""
    You are a skilled comic book writer.

    TASK:
    Generate a short comic book plot based on the story idea provided below. Also generate a description of the main
    character.
    Generate one sentence per scene, separated by periods. The story should be 3-7 sentences long. 
    IMPORTANT: Do NOT include any commentary, notes, or additional thoughts. Only output the story sentences and character description exactly as requested.

    Your output must include:
    - Story plot with one sentence per scene.
    - Very short description of the main character's appearance.
    - IMPORTANT!!! ALWAYS use a delimiter '---' to separate the story from the character description.

    STORY PROMPT: {prompt}
    """

    response = client.text_generation(
        model="mistralai/Mistral-7B-Instruct-v0.3",
        prompt=instructions,
        max_new_tokens=250,
        temperature=0.7,
    )
    return response


def remove_think_block(text: str):
    return re.sub(r'<think>.*?</think>', '', text, flags=re.DOTALL).strip()


def parse_screenwriter_output(output: str):
    cleaned_output = remove_think_block(output)
    delimiter = '---'
    if delimiter in cleaned_output:
        story, character = cleaned_output.split(delimiter, 1)
        return story.strip(), character.strip()
    else:
        lines = [line.strip() for line in cleaned_output.strip().split('\n') if line.strip()]
        if len(lines) < 2:
            return '', ''
        story = ' '.join(lines[:-1])
        character = lines[-1]
        return story, character


def error_image(message):
    img = Image.new("RGB", (512, 512), color=(255, 255, 255))
    d = ImageDraw.Draw(img)
    d.text((10, 250), message, fill=(255, 0, 0))
    return img


def illustrator(story: str, character: str):
    if not story or not character:
        raise ValueError('Could not parse story or character from input.')

    scenes = [s.strip() for s in story.split('.') if s.strip()]
    images = []

    for idx, scene in enumerate(scenes):
        prompt = f"Comic book style illustration. No text. Scene: {scene}. Character: {character}"
        try:
            image = pipe(prompt).images[0]
            images.append((image, scene))
        except Exception as e:
            images.append((error_image(f'Error: {str(e)}'), f'Error in scene {idx + 1}'))
    return images


def comic_pipeline(prompt: str):
    output = screenwriter(prompt)
    story, character = parse_screenwriter_output(output)
    if not story or not character:
        return output, [(error_image("Parse error: Could not extract story or character."), 'Parse error')]
    images = illustrator(story, character)
    return f"{story}\n---\n{character}", images


with gr.Blocks(theme=gr.themes.Ocean(), title='Comic Generator') as demo:
    gr.Markdown("# Comic Generator\nGive a prompt and get a comic!")
    with gr.Row():
        story_input = gr.Textbox(label='Story Prompt', placeholder='A unicorn named Jeff discovers a mysterious dish')
    generate_btn = gr.Button('Generate Comic')
    with gr.Row():
        story_output = gr.Textbox(label='Screenwriter Output', lines=6)
        gallery = gr.Gallery(label='Comic Scenes')
    generate_btn.click(comic_pipeline, inputs=story_input, outputs=[story_output, gallery])


if __name__ == "__main__":
    demo.launch()