import streamlit as st
from PIL import Image
from transformers import pipeline
from gtts import gTTS

st.set_page_config(page_title="Image to Audio Story", page_icon="🦜")
caption_pipeline = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
story_pipeline = pipeline("text-generation", model="Qwen/Qwen2.5-0.5B-Instruct")


def extract_image_caption(image_data):
    img_obj = Image.open(image_data)
    caption_results = caption_pipeline(img_obj)
    return caption_results[0]['generated_text']


def compose_story_from_caption(caption_detail):
    prompt_text = (
        "You are a talented and imaginative storyteller for children aged 3 to 10. "
        "Using the details derived from the image below, craft a captivating tale that includes three main characters, "
        "an adventurous journey, and delightful surprises. "
        "Your story should be vivid, original, and between 90 and 100 words in length.\n\n"
        f"Image Details: {caption_detail}\n\nStory:"
    )
    
    # 生成故事，设置最大长度为150，以确保有足够空间生成90-100词的内容
    story_results = story_pipeline(prompt_text, num_return_sequences=1, max_length=150)
    story_text = story_results[0]['generated_text']
    
    # 提取故事文本
    if "Story:" in story_text:
        return story_text.split("Story:", 1)[1].strip()
    return story_text.strip()
    
def convert_text_to_audio(text_content, audio_path="output.mp3"):
    """
    将文本转换为音频文件。
    """
    tts_engine = gTTS(text=text_content, lang="en")
    tts_engine.save(audio_path)
    return audio_path


def run_app():
    st.markdown("<h1 style='text-align: center;'>Your Image to Audio Story 🦜</h1>", unsafe_allow_html=True)
    st.write("Upload an image below and we will generate an engaging story from the picture, then convert the story into an audio playback!")

    uploaded_image = st.file_uploader("Select an Image", type=["png", "jpg", "jpeg"])

    if uploaded_image is not None:
        image_display = Image.open(uploaded_image)
        st.image(image_display, caption="Uploaded Image", use_container_width=True)
        
        with st.spinner("Generating caption for the image..."):
            caption_text = extract_image_caption(uploaded_image)
        st.write("**Generated Caption:**", caption_text)
        
        with st.spinner("Composing story..."):
            story_text = compose_story_from_caption(caption_text)
        st.write("**Story:**")
        st.write(story_text)
        
        with st.spinner("Converting text to audio..."):
            audio_file = convert_text_to_audio(story_text)
        st.audio(audio_file, format="audio/mp3")


if __name__ == "__main__":
    run_app()