app.py

import gradio as gr
from gtts import gTTS
import json

def generate_dialogue(rounds: int, method: str, role1: str, role2: str) -> list:
    roles = [role1, role2]
    dialogue = []
    for i in range(rounds):
        if method == "auto":
            dialogue.append((roles[i % 2], f"自動文本 {i + 1}"))
        else:
            dialogue.append((roles[i % 2], ""))
    return dialogue

def main_function(rounds: int, method: str, role1: str, role2: str):
    dialogue = generate_dialogue(rounds, method, role1, role2)
    
    audio_path = dialogue_to_audio(dialogue)
    json_output = json.dumps({"dialogue": dialogue}, ensure_ascii=False, indent=4)
    
    # 儲存對話為 JSON 文件
    with open("dialogue_output.json", "w", encoding="utf-8") as f:
        f.write(json_output)

    # 將對話格式化為 Chatbot 能接受的格式
    formatted_dialogue = [{"role": item[0], "content": item[1]} for item in dialogue]
    
    return formatted_dialogue, audio_path, "dialogue_output.json"

def dialogue_to_audio(dialogue):
    text = " ".join([item[1] for item in dialogue])
    tts = gTTS(text=text, lang='en')
    file_path = "temp_audio.mp3"
    tts.save(file_path)
    return file_path

if __name__ == "__main__":
    gr.Interface(
        main_function, 
        [
            gr.components.Slider(minimum=2, maximum=6, step=2, default=2, label="對話輪數"),
            gr.components.Dropdown(choices=["auto", "manual"], label="生成方式"),
            gr.components.Textbox(default="A", label="角色 1 名稱"),
            gr.components.Textbox(default="B", label="角色 2 名稱"),
        ], 
        [
            gr.components.Chatbot(label="生成的對話"),
            gr.components.Audio(type="filepath", label="對話朗讀"),
            gr.components.File(label="下載對話 JSON 文件")
        ]
    ).launch()