当前位置：首页 > news >正文

cpu运行 kokoro tts 服务器语音转化首选

news 2025/8/2 16:30:24

处理一些简单的语音不在话下，关键可以cpu 模式运行，也挺快的，实测占用内存大概2g ，意味着再一台 4c4g 的服务器就可以跑了，初始化需要下载

# pip install kokoro>=0.8.1 "misaki[zh]>=0.8.1"

在这里插入图片描述

from flask import Flask, request, jsonify, send_file
from kokoro import KModel, KPipeline
from pathlib import Path
import numpy as np
import soundfile as sf
import torch
import io
import base64app = Flask(__name__)# 初始化模型（在应用启动时只执行一次）
REPO_ID = 'hexgrad/Kokoro-82M-v1.1-zh'
SAMPLE_RATE = 24000
VOICE = 'zf_005'def init_model():global model, zh_pipeline# 设备配置device = 'cuda' if torch.cuda.is_available() else 'cpu'# 初始化模型和管道model = KModel(repo_id=REPO_ID).to(device).eval()# 定义英文处理函数en_pipeline = KPipeline(lang_code='a', repo_id=REPO_ID, model=False)def en_callable(text):if text == 'Kokoro':return 'kˈOkəɹO'elif text == 'Sol':return 'sˈOl'try:return next(en_pipeline(text)).phonemesexcept:return text  # 如果处理失败，返回原始文本zh_pipeline = KPipeline(lang_code='z', repo_id=REPO_ID, model=model, en_callable=en_callable)# 初始化模型
init_model()# HACK: Mitigate rushing caused by lack of training data beyond ~100 tokens
# Simple piecewise linear fn that decreases speed as len_ps increases
def speed_callable(len_ps):speed = 0.8if len_ps <= 83:speed = 1elif len_ps < 183:speed = 1 - (len_ps - 83) / 500return speed * 1.5def generate_tts(text):"""生成TTS音频的通用函数"""try:# 生成语音generator = zh_pipeline(text, voice=VOICE, speed=speed_callable)result = next(generator)wav = result.audio# 将 numpy 数组转换为 WAV 格式的字节流buffer = io.BytesIO()sf.write(buffer, wav, SAMPLE_RATE, format='WAV')buffer.seek(0)return bufferexcept Exception as e:raise Exception(f"TTS生成失败: {str(e)}")@app.route('/tts', methods=['POST'])
def text_to_speech():try:# 获取请求数据data = request.get_json()if not data or 'text' not in data:return jsonify({'error': 'Missing text parameter'}), 400text = data['text']if not text:return jsonify({'error': 'Text cannot be empty'}), 400buffer = generate_tts(text)# 返回音频文件return send_file(buffer,mimetype='audio/wav',as_attachment=True,download_name='output.wav')except Exception as e:return jsonify({'error': str(e)}), 500@app.route('/tts_get')
def text_to_speech_get():try:# 从查询参数获取文本text = request.args.get('text', '')if not text:return jsonify({'error': 'Missing text parameter'}), 400buffer = generate_tts(text)# 返回音频文件return send_file(buffer,mimetype='audio/wav',as_attachment=True,download_name='output.wav')except Exception as e:return jsonify({'error': str(e)}), 500@app.route('/tts_base64', methods=['POST'])
def text_to_speech_base64():try:# 获取请求数据data = request.get_json()if not data or 'text' not in data:return jsonify({'error': 'Missing text parameter'}), 400text = data['text']if not text:return jsonify({'error': 'Text cannot be empty'}), 400buffer = generate_tts(text)# 转换为 base64 编码wav_base64 = base64.b64encode(buffer.read()).decode('utf-8')# 返回 base64 编码的音频return jsonify({'status': 'success','audio': wav_base64,'format': 'wav'})except Exception as e:return jsonify({'error': str(e)}), 500@app.route('/tts_base64_get')
def text_to_speech_base64_get():try:# 从查询参数获取文本text = request.args.get('text', '')if not text:return jsonify({'error': 'Missing text parameter'}), 400buffer = generate_tts(text)# 转换为 base64 编码wav_base64 = base64.b64encode(buffer.read()).decode('utf-8')# 返回 base64 编码的音频return jsonify({'status': 'success','audio': wav_base64,'format': 'wav'})except Exception as e:return jsonify({'error': str(e)}), 500@app.route('/health', methods=['GET'])
def health_check():return jsonify({'status': 'healthy'})if __name__ == '__main__':# 禁用调试模式以避免重新加载问题app.run(host='0.0.0.0', port=5000, debug=False)