import streamlit as st
from transformers import pipeline

# function part
# img2text
def img2text(url):
    image_to_text_model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
    text = image_to_text_model(url)[0]["generated_text"]
    return text

# text2story
def text2story(text):
    story_generator = pipeline("text-generation", model="aspis/gpt2-genre-story-generation")
    story_text = story_generator(text, max_length=150, num_return_sequences=1)
    return story_text[0]["generated_text"]

# text2audio
def text2audio(story_text):
    tts_model = pipeline("text-to-speech", model="facebook/mms-tts-eng")
    audio_data = tts_model(story_text)
    return audio_data


# Main part
def main():
    st.set_page_config(page_title="Your Image to Audio Story", page_icon="🦜")
    st.header("Turn Your Image to Audio Story")

    if "scenario" not in st.session_state:
        st.session_state.scenario = None
    if "story" not in st.session_state:
        st.session_state.story = None
    if "audio_data" not in st.session_state:
        st.session_state.audio_data = None

    uploaded_file = st.file_uploader("Select an Image...")

    if uploaded_file is not None and st.session_state.scenario is None:
        print(uploaded_file)
        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_container_width=True)

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

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

        # Stage 3: Story to Audio Data
        st.text('Generating audio data...')
        st.session_state.audio_data = text2audio(st.session_state.story)

    elif st.session_state.scenario:
        st.image(uploaded_file, caption="Uploaded Image", use_container_width=True)
        st.write("Image Caption: ", st.session_state.scenario)
        st.write("Generated Story: ", st.session_state.story)

    # Play button (No reprocessing)
    if st.session_state.audio_data and st.button("Play Audio"):
        st.audio(st.session_state.audio_data['audio'], 
                 format="audio/wav", 
                 start_time=0, 
                 sample_rate=st.session_state.audio_data['sampling_rate'])

if __name__ == "__main__":
    main()