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

st.set_page_config(page_title="Image to Audio Story", page_icon="🦜")

def extract_image_caption(image_data):
"""
利用预训练模型从图像中提取描述性文字。
"""
img_obj = Image.open(image_data)
caption_pipeline = pipeline(
"image-to-text",
model="Salesforce/blip-image-captioning-base",
)
caption_results = caption_pipeline(img_obj)
caption_text = caption_results[0]['generated_text']
return caption_text

def compose_story_from_caption(caption_detail):
"""
根据图像描述创作一篇充满创意的儿童故事。
"""
story_pipeline = pipeline(
"text-generation",
model="Qwen/Qwen2-1.5B",
)
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 goes beyond merely describing the scene. "
"Let your creativity shine by introducing engaging characters, adventurous journeys, and delightful surprises. "
"Your story should be vivid, original, and between 100 and 300 words in length.\n\n"
f"Image Details: {caption_detail}\n\nStory:"
)
story_results = story_pipeline(prompt_text, num_return_sequences=1)
story_text = story_results[0]['generated_text']
if "Story:" in story_text:
story = story_text.split("Story:", 1)[1].strip()
else:
story = story_text.strip()
return story

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()