app.py

import gradio as gr
import easyocr
from deep_translator import GoogleTranslator
from gtts import gTTS
import numpy as np
import os

# --- 全局初始化 ---
# 确保在 CPU 模式下运行，并使用更稳定的 deep-translator 库
OCR_READER = None

def get_ocr_reader():
    """初始化 EasyOCR Reader，使用 CPU 模式和轻量化模型，避免内存溢出 (OOM)"""
    global OCR_READER
    if OCR_READER is None:
        try:
            # 使用 CPU 模式 (gpu=False)
            # 第一次运行时，会下载模型文件到本地的 model_cache 目录
            OCR_READER = easyocr.Reader(['en'], 
                                       gpu=False, 
                                       model_storage_directory='./model_cache', 
                                       download_enabled=True)
            print("EasyOCR Reader 初始化成功 (CPU 模式)")
        except Exception as e:
            print(f"EasyOCR 初始化失败: {e}")
            raise
    return OCR_READER

def process_image(image):
    """主处理函数：OCR -> 翻译 -> TTS"""
    
    # 检查输入
    if image is None:
        return "请拍摄或上传图片", None
    
    reader = get_ocr_reader()
    
    # 1. 文本识别 (OCR)
    try:
        results = reader.readtext(image, detail=0)
        source_text = " ".join(results).strip()
    except Exception as e:
        return f"文字识别失败: {str(e)}", None

    if not source_text:
        return "未能识别到图片中的文字。", None

    # 2. 翻译 (使用 deep-translator，稳定且无需密钥)
    try:
        # 从自动检测语言翻译到中文 (zh-CN)
        translated_text = GoogleTranslator(source='auto', target='zh-CN').translate(source_text)
    except Exception as e:
        return f"翻译器暂时不可用: {str(e)}", None

    # 3. 语音播报 (TTS)
    audio_path = "output.mp3"
    try:
        tts = gTTS(text=translated_text, lang='zh-cn')
        tts.save(audio_path)
    except Exception as e:
        print(f"语音合成失败: {e}")
        audio_path = None

    result_text = f"【原文】:\n{source_text}\n\n【翻译】:\n{translated_text}"
    return result_text, audio_path

# --- Gradio 界面设计 (解决 theme 和 sources 兼容性问题) ---
with gr.Blocks() as demo:
    gr.Markdown("# 📸 拍照翻译官 (最终稳定版)")
    gr.Markdown("对准英文文字拍照，稍等片刻即可获得中文翻译及发音。")
    
    with gr.Row():
        with gr.Column():
            # 兼容性修正：将 'camera' 替换为旧版 Gradio 兼容的 'webcam'
            input_img = gr.Image(sources=["webcam", "upload"], type="numpy", label="上传/拍照")
            btn = gr.Button("开始翻译", variant="primary")
        with gr.Column():
            out_txt = gr.Textbox(label="文本结果", lines=8)
            out_audio = gr.Audio(label="语音播报")
    
    btn.click(process_image, input_img, [out_txt, out_audio])

if __name__ == "__main__":
    demo.launch()