app.py

import gradio as gr

# Using openai models ---------------------------------------------------------

from langchain_openai import OpenAI
import os
openai_api_key = os.getenv("OPENAI_API_KEY")

import io
import base64
import requests
import json

width = 800


# Function to call the API for image and get the response
def get_response_for_image(openai_api_key, image):
    base64_image = base64.b64encode(image).decode('utf-8')
    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {openai_api_key}"
    }
    payload = {
        "model": "gpt-4o",
        "messages": [
          {
            "role": "user",
            "content": [
              {
                "type": "text",
                "text": '''Describe or caption the image within 20 words. Output in json format with key: Description'''
              },
              {
                "type": "image_url",
                "image_url": {
                  "url": f"data:image/jpeg;base64,{base64_image}",
                  "detail": "low"
                }
              }
            ]
          }
        ],
        "max_tokens": 200
    }
    response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)
    return response.json()


def generate_story(image, theme, genre, word_count):
    try:
        # Convert PIL image to bytes-like format
        with io.BytesIO() as output:
            image.save(output, format="JPEG")
            image_bytes = output.getvalue()
        
        # Decode the caption
        caption_response = get_response_for_image(openai_api_key, image_bytes)
        json_str = caption_response['choices'][0]['message']['content']
        json_str = json_str.replace('```json', '').replace('```', '').strip()
        content_json = json.loads(json_str)
        caption_text = content_json['Description']
        
        
        # Generate story based on the caption
        story_prompt = f"Write an interesting {theme} story in the {genre} genre about {caption_text}. The story should be within {word_count} words."

        llm = OpenAI(model_name="gpt-3.5-turbo-instruct", openai_api_key=openai_api_key, max_tokens=1000)
        story = llm.invoke(story_prompt)
        
        return caption_text, story
    except Exception as e:
        return f"An error occurred during inference: {str(e)}"
    
    
# Using open source models ----------------------------------------------------

'''
from transformers import pipeline, AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel 

# Load text generation model

text_generation_model = pipeline("text-generation", model="distilbert/distilgpt2")

# Load image captioning model

encoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"
decoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"
model_checkpoint = "nlpconnect/vit-gpt2-image-captioning"

feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)
model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint)

def generate_story(image, theme, genre, word_count):
    try:
        # Preprocess the image
        image = image.convert('RGB')
        image_features = feature_extractor(images=image, return_tensors="pt")
        
        # Generate image caption
        caption_ids = model.generate(image_features.pixel_values, max_length=50, num_beams=3, temperature=1.0)
        
        # Decode the caption
        caption_text = tokenizer.batch_decode(caption_ids, skip_special_tokens=True)[0]
        
        # Generate story based on the caption
        story_prompt = f"Write an interesting {theme} story in the {genre} genre. The story should be within {word_count} words about {caption_text}."
        story = text_generation_model(story_prompt, max_length=150)[0]["generated_text"]
        
        return caption_text, story
        
    except Exception as e:
        return f"An error occurred during inference: {str(e)}"
'''


# ------------------------------------------------------------------------- 

# Gradio interface
input_image = gr.Image(label="Select Image",type="pil")
input_theme = gr.Dropdown(["Love and Loss", "Identity and Self-Discovery", "Power and Corruption", "Redemption and Forgiveness", "Survival and Resilience", "Nature and the Environment", "Justice and Injustice", "Friendship and Loyalty", "Hope and Despair"], label="Input Theme")
input_genre = gr.Dropdown(["Fantasy", "Science Fiction", "Poetry", "Mystery/Thriller", "Romance", "Historical Fiction", "Horror", "Adventure", "Drama", "Comedy"], label="Input Genre")
output_caption = gr.Textbox(label="Image Caption", lines=3)
output_text = gr.Textbox(label="Generated Story",lines=20)
examples = [
    ["example1.jpg", "Love and Loss", "Fantasy", 80],
    ["example2.jpg", "Identity and Self-Discovery", "Science Fiction", 100],
    ["example3.jpg", "Power and Corruption", "Mystery/Thriller", 120],
    ["example4.jpg", "Redemption and Forgiveness", "Romance", 80],
    ["example5.jpg", "Survival and Resilience", "Poetry", 150],
    ["example6.jpg", "Nature and the Environment", "Horror", 120],
    ["example7.jpg", "Justice and Injustice", "Adventure", 80],
    ["example8.jpg", "Friendship and Loyalty", "Drama", 100],
]
word_count_slider = gr.Slider(minimum=50, maximum=200, value=80, step=5, label="Word Count")


gr.Interface(
    fn=generate_story,
    inputs=[input_image, input_theme, input_genre, word_count_slider],
    theme='freddyaboulton/dracula_revamped',
    outputs=[output_caption, output_text],
    examples = examples,
    title="Image to Story Generator",
    description="Generate a story from an image taking theme and genre as input. It leverages image captioning and text generation models.",
).launch()