app.py

import streamlit as st
from transformers import pipeline, AutoProcessor, AutoModel
from scipy.io.wavfile import write as write_wav
import numpy as np
import torch

def img2text(image_path):
    img2caption = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
    return img2caption(image_path)[0]['generated_text']

def text2story(text):
    pipe = pipeline("text-generation", model="pranavpsv/genre-story-generator-v2")
    story_text = pipe(text, max_length=100)[0]['generated_text']
    return story_text

def text2audio(story_text):
    processor = AutoProcessor.from_pretrained("facebook/mms-tts-eng")
    model = AutoModel.from_pretrained("facebook/mms-tts-eng")

    inputs = processor(text=story_text, return_tensors="pt")
    with torch.no_grad():
        output = model(**inputs).waveform

    audio_array = output.cpu().numpy().squeeze()
    sample_rate = 16000 

    return audio_array, sample_rate

# Streamlit UI
st.set_page_config(page_title="Your Image to Audio Story", page_icon="🦜")
st.header("Turn Your Image to Audio Story")
uploaded_file = st.file_uploader("Select an Image...")

if uploaded_file is not None:
    bytes_data = uploaded_file.getvalue()
    with open(uploaded_file.name, "wb") as file:
        file.write(bytes_data)
    st.image(uploaded_file, caption="Uploaded Image", use_column_width=True)

    # Stage 1: Image to Text
    st.text('Processing img2text...')
    scenario = img2text(uploaded_file.name)
    st.write(scenario)

    # Stage 2: Text to Story
    st.text('Generating a story...')
    story = text2story(scenario)
    st.write(story)

    # Stage 3: Story to Audio
    st.text('Generating audio...')
    audio_array, sample_rate = text2audio(story)
    
    audio_file = "output_audio.wav"
    write_wav(audio_file, sample_rate, audio_array)
    st.audio(audio_file)