#!/usr/bin/env python3 """MiMo V2.5 TTS — 独立语音合成脚本 用法: # 基本用法 python3 mimo_tts.py "要合成的文字" # 指定音色和输出路径 python3 mimo_tts.py "Hello world" --voice Mia --output hello.wav # 使用语音设计模型（通过文字描述生成音色） python3 mimo_tts.py "Once upon a time" --model voicedesign \ --style "A warm storyteller voice with British accent" # 列出所有可用音色 python3 mimo_tts.py --list-voices # 流式输出（PCM16 24kHz） python3 mimo_tts.py "你好世界" --stream --output output.wav 环境变量: XIAOMI_API_KEY — 必需，小米 MiMo API Key XIAOMI_BASE_URL — 可选，API 端点（默认 https://api.xiaomimimo.com/v1） 获取 API Key: https://mimo.mi.com """ from __future__ import annotations import argparse import base64 import json import os import sys import urllib.request from pathlib import Path # --------------------------------------------------------------------------- # Constants # --------------------------------------------------------------------------- API_BASE_DEFAULT = "https://api.xiaomimimo.com/v1" VOICES = [ {"id": "mimo_default", "display": "默认 (auto)", "lang": "auto", "gender": "-"}, {"id": "冰糖", "display": "冰糖 Bingtang", "lang": "zh", "gender": "♀"}, {"id": "茉莉", "display": "茉莉 Moli", "lang": "zh", "gender": "♀"}, {"id": "苏打", "display": "苏打 Suda", "lang": "zh", "gender": "♂"}, {"id": "白桦", "display": "白桦 Baihua", "lang": "zh", "gender": "♂"}, {"id": "Mia", "display": "Mia", "lang": "en", "gender": "♀"}, {"id": "Chloe", "display": "Chloe", "lang": "en", "gender": "♀"}, {"id": "Milo", "display": "Milo", "lang": "en", "gender": "♂"}, {"id": "Dean", "display": "Dean", "lang": "en", "gender": "♂"}, ] MODELS = { "tts": "mimo-v2.5-tts", "voicedesign": "mimo-v2.5-tts-voicedesign", "voiceclone": "mimo-v2.5-tts-voiceclone", } # --------------------------------------------------------------------------- # Core API # --------------------------------------------------------------------------- def get_api_base() -> str: return os.environ.get("XIAOMI_BASE_URL", API_BASE_DEFAULT).rstrip("/") def synthesize( text: str, output_path: str, *, voice: str = "mimo_default", model: str = "mimo-v2.5-tts", style: str | None = None, api_key: str | None = None, ) -> str: """合成语音并写入文件，返回文件路径。""" api_key = api_key or os.environ.get("XIAOMI_API_KEY") if not api_key: raise RuntimeError("XIAOMI_API_KEY is not set. Get one at https://mimo.mi.com") messages: list = [] if style: messages.append({"role": "user", "content": style}) messages.append({"role": "assistant", "content": text}) audio_config: dict = {"format": "wav"} if "voicedesign" not in model and "voiceclone" not in model: audio_config["voice"] = voice body = { "model": model, "messages": messages, "audio": audio_config, } data = json.dumps(body).encode() req = urllib.request.Request( f"{get_api_base()}/chat/completions", data=data, headers={"Content-Type": "application/json", "api-key": api_key}, method="POST", ) with urllib.request.urlopen(req, timeout=60) as resp: resp_data = json.loads(resp.read().decode()) audio_b64 = resp_data["choices"][0]["message"]["audio"]["data"] audio_bytes = base64.b64decode(audio_b64) output_path = str(Path(output_path).expanduser()) if not output_path.endswith(".wav"): output_path = os.path.splitext(output_path)[0] + ".wav" os.makedirs(os.path.dirname(output_path) or ".", exist_ok=True) with open(output_path, "wb") as f: f.write(audio_bytes) return output_path def synthesize_stream( text: str, *, voice: str = "mimo_default", model: str = "mimo-v2.5-tts", style: str | None = None, api_key: str | None = None, ): """流式合成，yield PCM16 字节块 (24kHz mono)。""" api_key = api_key or os.environ.get("XIAOMI_API_KEY") if not api_key: raise RuntimeError("XIAOMI_API_KEY is not set.") messages: list = [] if style: messages.append({"role": "user", "content": style}) messages.append({"role": "assistant", "content": text}) audio_config: dict = {"format": "pcm16"} if "voicedesign" not in model and "voiceclone" not in model: audio_config["voice"] = voice body = {"model": model, "messages": messages, "audio": audio_config, "stream": True} data = json.dumps(body).encode() req = urllib.request.Request( f"{get_api_base()}/chat/completions", data=data, headers={"Content-Type": "application/json", "api-key": api_key}, method="POST", ) with urllib.request.urlopen(req, timeout=120) as resp: for line in resp: line = line.strip() if not line or not line.startswith(b"data: "): continue payload = line[6:] if payload == b"[DONE]": break try: chunk = json.loads(payload) audio_data = chunk["choices"][0]["delta"].get("audio", {}) if isinstance(audio_data, dict) and "data" in audio_data: yield base64.b64decode(audio_data["data"]) except (json.JSONDecodeError, KeyError, IndexError): continue # --------------------------------------------------------------------------- # CLI # --------------------------------------------------------------------------- def list_voices(): print(f"{'ID':<15} {'Display':<20} {'Lang':<6} {'Gender'}") print("-" * 50) for v in VOICES: print(f"{v['id']:<15} {v['display']:<20} {v['lang']:<6} {v['gender']}") def main(): parser = argparse.ArgumentParser( description="MiMo V2.5 TTS — 语音合成", formatter_class=argparse.RawDescriptionHelpFormatter, epilog=__doc__, ) parser.add_argument("text", nargs="?", help="要合成的文字") parser.add_argument("-v", "--voice", default="mimo_default", help="音色 ID (默认: mimo_default)") parser.add_argument("-m", "--model", default="tts", choices=["tts", "voicedesign", "voiceclone"], help="模型 (默认: tts)") parser.add_argument("-s", "--style", default=None, help="风格/语气描述 (user message)") parser.add_argument("-o", "--output", default=None, help="输出文件路径 (默认: ~/mimo_tts_output.wav)") parser.add_argument("--stream", action="store_true", help="流式模式 (输出 PCM16 WAV)") parser.add_argument("--list-voices", action="store_true", help="列出所有可用音色") parser.add_argument("--list-models", action="store_true", help="列出所有可用模型") args = parser.parse_args() if args.list_voices: list_voices() return if args.list_models: for key, mid in MODELS.items(): print(f" {key:<15} → {mid}") return if not args.text: parser.error("请提供要合成的文字") model_id = MODELS[args.model] output = args.output or os.path.expanduser("~/mimo_tts_output.wav") if args.stream: print(f"流式合成中... voice={args.voice} model={model_id}") collected = bytearray() for chunk in synthesize_stream(args.text, voice=args.voice, model=model_id, style=args.style): collected.extend(chunk) print(f" 收到 {len(chunk)} 字节", end="\r") # Write as WAV (PCM16 24kHz mono) import struct, wave output = output if output.endswith(".wav") else output + ".wav" with wave.open(output, "wb") as wf: wf.setnchannels(1) wf.setsampwidth(2) wf.setframerate(24000) wf.writeframes(bytes(collected)) print(f"\n✅ 已保存: {output} ({len(collected)} 字节 PCM)") else: print(f"合成中... voice={args.voice} model={model_id}") result = synthesize(args.text, output, voice=args.voice, model=model_id, style=args.style) size_kb = os.path.getsize(result) / 1024 print(f"✅ 已保存: {result} ({size_kb:.1f} KB)") if __name__ == "__main__": main()