import gradio as gr import edge_tts import aiohttp import asyncio import tempfile, time from typing import TypedDict, List, Dict from pathlib import Path class VoiceTag(TypedDict, total=False): ContentCategories: List[str] VoicePersonalities: List[str] class VoiceMetadata(TypedDict, total=False): Name: str ShortName: str Gender: str Locale: str SuggestedCodec: str FriendlyName: str Status: str VoiceTag: VoiceTag async def list_voices() -> Dict[str, VoiceMetadata]: """ Fetch available voices from Microsoft Edge TTS. Returns: Dict[str, VoiceMetadata]: A dictionary mapping display names (e.g. "en-US-JennyNeural - Female") to their corresponding in the form like { 'Name': 'Microsoft Server Speech Text to Speech Voice (zh-CN, XiaoxiaoNeural)', 'ShortName': 'zh-CN-XiaoxiaoNeural', 'Gender': 'Female', 'Locale': 'zh-CN', 'SuggestedCodec': 'audio-24khz-48kbitrate-mono-mp3', 'FriendlyName': 'Microsoft Xiaoxiao Online (Natural) - Chinese (Mainland)', 'Status': 'GA', 'VoiceTag': { 'ContentCategories': ['News', 'Novel'], 'VoicePersonalities': ['Warm'] } } """ priority_order = [ "zh", # 中文优先 "en-US", # 美式英语 "en-GB", # 英式英语 "en", # 其他英语 "hi", # 印地语 "es", # 西班牙语 "ar", # 现代标准阿拉伯语 "fr", # 法语 "bn", # 孟加拉语 "pt", # 葡萄牙语 "ru", # 俄语 "id", # 印尼语 "ur", # 乌尔都语 "de", # 德语 "ja", # 日语 "pcm", # 尼日利亚皮钦语 "ar-EG", # 埃及阿拉伯语 "mr", # 马拉地语 "vi", # 越南语 "te", # 泰卢固语 "ha", # 豪萨语 "tr", # 土耳其语 "pa", # 西旁遮普语 "sw", # 斯瓦希里语 "fil", # 他加禄语 "ta", # 泰米尔语 "yue", # 粤语 "wuu", # 吴语 "fa", # 波斯语 "ko", # 韩语 "th", # 泰语 "jv", # 爪哇语 ] lang_priority = {lang: i for i, lang in enumerate(priority_order)} def get_priority(short_name: str) -> tuple[int, str]: """ Return (priority_rank, short_name) for stable sorting. """ for prefix, rank in lang_priority.items(): if short_name.startswith(prefix): return (rank, short_name) return (len(priority_order), short_name) # 其他语言排最后，内部按 short_name n = 5 for i in range(n): try: voices = await edge_tts.list_voices() break except Exception as e: time.sleep(1 * i) print("Retrying due to handshake error:", e) if i == n - 1: raise e # ✅ 先按优先级，再按 short_name 排序，保证同类不乱 voices_sorted = sorted(voices, key=lambda v: get_priority(v["ShortName"])) return {f"{v['ShortName']} - {v['Gender']}": v for v in voices_sorted} async def text_to_speech(text: str, voice: str, rate: int = 0, pitch: int = 0) -> str: """ Convert input text to speech using Microsoft Edge TTS. Args: text (str): The text to synthesize into speech. voice (str): The selected voice in the format "ShortName - Gender". rate (int, optional): Speech rate adjustment percentage. Default is 0. pitch (int, optional): Pitch adjustment in Hz. Default is 0. Returns: str: Path to the generated MP3 file. """ voice_short_name = voice.split(" - ")[0] rate_str = f"{rate:+d}%" pitch_str = f"{pitch:+d}Hz" n = 5 for i in range(n): try: communicate = edge_tts.Communicate( text, voice_short_name, rate=rate_str, pitch=pitch_str ) break except aiohttp.client_exceptions.WSServerHandshakeError as e: time.sleep(1 * i) print("Retrying due to handshake error:", e) if i == n - 1: raise e with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file: tmp_path = tmp_file.name await communicate.save(tmp_path) path = Path(tmp_path) return path.as_posix()#后面output到gr.Audio上，通过程序交互的时候gradio会将其转化为url，很神奇！ i18n = gr.I18n( en={"pageTitle": "# 🎙️ Microsoft Online Edge TTS & MCP Server", "generateSpeech": "Generate Speech", 'description':"""Convert text to speech using the free [Microsoft Edge TTS](https://github.com/rany2/edge-tts), API and MCP are available. Made by [MathJoy](https://www.cnblogs.com/imath).""", "selectVoice":"Select Voice", "speechRateAdjustment":"Speech Rate Adjustment (%)", "pitchAdjustment":"Pitch Adjustment (Hz)", "inputText":"Input Text" }, zh={"pageTitle": "# 🎙️ 微软线上 Edge TTS & MCP", "generateSpeech": "生成语音", 'description':"""文本语音合成，基于免费的[Microsoft Edge TTS](https://github.com/rany2/edge-tts)，由[MathJoy](https://www.cnblogs.com/imath)精心制作，也可通过API或MCP来使用。""", "selectVoice":"选择音色", "speechRateAdjustment":"语速调整 (%)", "pitchAdjustment":"音量调整 (Hz)", "inputText":"输入文字" } ) async def create_UI(): voices = await list_voices() with gr.Blocks(title="Microsoft Edge TTS & MCP",analytics_enabled=False) as UI: gr.api(#默认只在fn处暴露工具，这里主动暴露这个tool list_voices ) gr.Markdown(i18n("pageTitle")) with gr.Row(): with gr.Column(scale=1): gr.Markdown(i18n("description")) with gr.Row(): with gr.Column(): text_input = gr.Textbox(label=i18n("inputText"), lines=5,value='Your text here') voiceList=list(voices.keys()) voice_dropdown = gr.Dropdown( choices=voiceList, label=i18n("selectVoice"), value=voiceList[0]) rate_slider = gr.Slider( minimum=-50, maximum=50, value=0, label=i18n("speechRateAdjustment"), step=1 ) pitch_slider = gr.Slider( minimum=-20, maximum=20, value=0, label=i18n("pitchAdjustment"), step=1 ) generate_btn = gr.Button(i18n("generateSpeech"), variant="primary") audio_output = gr.Audio(label="Generated Audio", type="filepath",autoplay=True) generate_btn.click( fn=text_to_speech, inputs=[text_input, voice_dropdown, rate_slider, pitch_slider], outputs=[audio_output] ) return UI async def main(): UI = await create_UI() UI.queue(default_concurrency_limit=50) UI.launch( # show_api=False, i18n=i18n, mcp_server=True # ✅ make this an MCP server ) if __name__ == "__main__": asyncio.run(main())