app.py

# Imports
import os
import streamlit as st
import requests
from transformers import pipeline
import openai

# Suppressing all warnings
import warnings
warnings.filterwarnings("ignore")

# Image-to-text
def img2txt(url):
    print("Initializing captioning model...")
    captioning_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
    
    print("Generating text from the image...")
    text = captioning_model(url, max_new_tokens=20)[0]["generated_text"]
    
    print("Text generated successfully.")
    return text

# Text-to-story
def txt2story(img_text):
    print("Initializing client...")
    client = openai.OpenAI(
        api_key=os.environ["TOGETHER_API_KEY"], 
        base_url='https://api.together.xyz',
    )

    print("Constructing prompt for story generation...")
    content_prompt = f'''Based on the image description "{img_text}", conclude the story. 
    Resolve the conflict or summarize the outcome of the situation. Ensure the story MUST
    have a definitive ending. The end.'''

    print("Preparing message sequences for interaction...")
    messages = [
        {"role": "system", "content": "Once upon a time..."},
        {"role": "user", "content": img_text, "temperature": 1},
        {"role": "system", 
         "content": content_prompt, 
         "temperature": 0.7},
    ]
    
    print("Generating story completion using the AI model...")
    chat_completion = client.chat.completions.create(
        messages=messages,
        model="mistralai/Mixtral-8x7B-Instruct-v0.1",
        max_tokens=200)

    print("Story generated successfully.")
    return chat_completion.choices[0].message.content


# Text-to-speech
def txt2speech(text):
    print("Initializing text-to-speech conversion...")
    API_URL = "https://api-inference.huggingface.co/models/espnet/kan-bayashi_ljspeech_vits"
    headers = {"Authorization": f"Bearer {os.environ['HUGGINGFACEHUB_API_TOKEN']}"}
    payloads = {'inputs': text}

    print("Sending request for speech synthesis...")
    response = requests.post(API_URL, headers=headers, json=payloads)
    
    print("Saving synthesized speech to audio file...")
    with open('audio_story.mp3', 'wb') as file:
        file.write(response.content)
    
    print("Text-to-speech conversion completed.")


# Streamlit web app main function
def main():
    st.set_page_config(page_title="🎨 Image-to-Audio Story 🎧", page_icon="🖼️")
    st.title("Turn the Image into Audio Story")

# Allows users to upload an image file
    uploaded_file = st.file_uploader("# 📷 Upload an image...", type=["jpg", "jpeg", "png"])

    if uploaded_file is not None:
# Reads and saves uploaded image file
        bytes_data = uploaded_file.read()
        with open("uploaded_image.jpg", "wb") as file:
            file.write(bytes_data)

        st.image(uploaded_file, caption='🖼️ Uploaded Image', use_column_width=True)

 # Initiates AI processing and story generation
        with st.spinner("## 🤖 AI is at Work! "):
            scenario = img2txt("uploaded_image.jpg")  # Extracts text from the image
            story = txt2story(scenario)  # Generates a story based on the image text
            txt2speech(story)  # Converts the story to audio

            st.markdown("---")
            st.markdown("## 📜 Image Caption")
            st.write(scenario)

            st.markdown("---")
            st.markdown("## 📖 Story")
            st.write(story)

            st.markdown("---")
            st.markdown("## 🎧 Audio Story")
            st.audio("audio_story.mp3")
             
if __name__ == '__main__':
    main()

# Credits
st.markdown("### Credits")
st.caption('''
            Made with ❤️ by @Aditya-Neural-Net-Ninja\n 
            Utilizes Image-to-text, Text Generation, Text-to-speech Transformer Models\n
            Gratitude to Streamlit, 🤗 Spaces for Deployment & Hosting
            ''')