app.py

import os
import streamlit as st
import requests
from transformers import pipeline
import openai
from langchain import LLMChain, PromptTemplate
from langchain import HuggingFaceHub

from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler
import torch
from diffusers import DiffusionPipeline



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

api_token = os.getenv('H_TOKEN')

# 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)
    return text

# Text-to-story

model = "tiiuae/falcon-7b-instruct"
llm = HuggingFaceHub(
    huggingfacehub_api_token = api_token,
    repo_id = model,
    verbose = False,
    model_kwargs = {"temperature":0.2, "max_new_tokens": 4000})

def generate_story(scenario, llm):
  template= """You are a story teller.
               You get a scenario as an input text, and generates a short story out of it.
               Context: {scenario}
               Story:
               """
  prompt = PromptTemplate(template=template, input_variables=["scenario"])
  #Let's create our LLM chain now
  chain = LLMChain(prompt=prompt, llm=llm)
  story = chain.predict(scenario=scenario)
  start_index = story.find("Story:") + len("Story:")

  # Extract the text after "Story:"
  story = story[start_index:].strip()
  return story


# 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 {api_token }"}
    payloads = {'inputs': text}

    response = requests.post(API_URL, headers=headers, json=payloads)
    
    with open('audio_story.mp3', 'wb') as file:
        file.write(response.content)


# text-to- image
def txt2img(text, style="realistic"):
    model_id = "stabilityai/stable-diffusion-2"
    #pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0")

    # Use the Euler scheduler here instead
    scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
    pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)
    pipe = pipe.to("cuda")
    image = pipe(prompt = text, guidance_scale = 7.5).images[0]
    return image
    
        
st.sidebar.title("Choose the task")
        
# Function for the Audio Story page
def audio_story_page():
    st.title("🎨 Image-to-Audio Story 🎧")
    st.write("Turn the Image into Audio Story")
    
    uploaded_file = st.file_uploader("Upload an image...", type=["jpg", "jpeg", "png"])

    if uploaded_file is not None:
        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)

        with st.spinner("AI is at Work!"):
            scenario = img2txt("uploaded_image.jpg")
            story = generate_story(scenario, llm)
            txt2speech(story)

            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")

# Function for the Image Generator page
def image_generator_page():
    st.title("Stable Diffusion Image Generation")
    st.write("This app lets you generate images using Stable Diffusion with the Euler scheduler.")
        
    prompt = st.text_input("Enter your prompt:")
    image_style = st.selectbox("Style Selection", ["realistic", "cartoon", "watercolor"])
        
    if st.button("Generate Image"):
        if prompt:
            with st.spinner("Generating image..."):
                image = txt2img(prompt, style=image_style)
            st.image(image)
        else:
            st.error("Please enter a prompt...")

# Function for the Home page
def home_page():
    st.title("Welcome to your Creative Canvas!")
    st.write("Use the tools in the sidebar to create audio stories and unique images.")

# Streamlit web app main function
def main():
    selection = st.sidebar.radio("Go to", ["Home", "Audio Story", "Image Generator"])

    if selection == "Home":
        home_page()
    elif selection == "Audio Story":
        audio_story_page()
    elif selection == "Image Generator":
        image_generator_page()

if __name__ == '__main__':
    main()