app.py

import gradio as gr
# from langchain.llms import OpenAI
from langchain_openai import OpenAI
from transformers import pipeline
from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel 

import os
openai_api_key = os.getenv("OPENAI_API_KEY")

# Load text generation model
# text_generation_model = pipeline("text-generation", model="openai-community/gpt2-large")
# 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):
    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 100 words about {caption_text}."

        llm = OpenAI(model_name="gpt-3.5-turbo-instruct", openai_api_key=openai_api_key)
        story = llm.invoke(story_prompt)
        # 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=2)
output_text = gr.Textbox(label="Generated Story",lines=8)
examples = [f"example{i}.jpg" for i in range(1,2)]


gr.Interface(
    fn=generate_story,
    inputs=[input_image, input_theme, input_genre],
    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()