app.py

import os
import textwrap
from PIL import Image, ImageDraw, ImageFont

import gradio as gr
import numpy as np
import torch
from lavis.models import load_model_and_preprocess
import openai

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

model, vis_processors, _ = load_model_and_preprocess(
    name="blip2_opt", model_type="pretrain_opt2.7b", is_eval=True, device=device
)

openai.api_key = os.environ["OPENAI_API_KEY"]

def generate_caption(image):
    pil_image = image.copy()  # Create a copy of the input PIL image
    image = vis_processors["eval"](image).unsqueeze(0).to(device)
    caption = model.generate({"image": image})

    caption = "\n".join(caption)
    #use gpt-4 to generate a meme based on the caption
    response = openai.ChatCompletion.create(
                model="gpt-4",
                messages=[
                    {"role": "system", "content": "Escribe un meme chistoso en español para una imagen a partir en la descripción dada por el usuario. No uses emojis. El meme tiene que ser corto y gracioso. El output del asistente solo debe ser el meme. Asegúrate que el meme sea tan bueno que se vuelva viral!"},
                    {"role": "user", "content": caption}
                ],
                temperature=0.6
            )
    
    meme_text = response.choices[0].message.content
    print(meme_text)

    # Put the meme text on the image
    draw = ImageDraw.Draw(pil_image)

    # Determine font size
    max_width = int(pil_image.width * 0.9)
    font_size = int(max_width / (len(meme_text) / 2))
    font = ImageFont.truetype("impact.ttf", font_size)

    # Wrap the text to fit within the image width and have a maximum of 2 lines
    wrapped_text = textwrap.fill(meme_text, width=int(max_width / font.getsize("A")[0]))
    text_lines = wrapped_text.split('\n')

    y = 10
    for line in text_lines:
        line_width, line_height = draw.textsize(line, font=font)
        x = (pil_image.width - line_width) // 2
        draw.text((x, y), line, fill=(255, 255, 255), font=font)
        y += line_height

    pil_image = pil_image.convert('RGB')

    if torch.cuda.is_available():
        torch.cuda.empty_cache()

    return pil_image


with gr.Blocks() as demo:
    gr.Markdown(
        "### Memero - Generador de Memes"
    )
    gr.Markdown(
        "Genera un meme en español a partir de una imagen."
    )

    with gr.Row():
        with gr.Column():
            input_image = gr.Image(label="Imagen", type="pil")
            btn_caption = gr.Button("Generar meme")
            output_text = gr.Image(label="Meme", lines=5)

    btn_caption.click(
        generate_caption, inputs=[input_image], outputs=[output_text]
    )

demo.launch()