用cosyvoice批量把生成有声书

CosyVoice是一个功能强大的多语言语音生成模型，由阿里巴巴的通义实验室（FunAudioLLM团队）开发。它不仅能将文本合成为高度拟人的自然语音，还具备零样本语音克隆、跨语言合成等前沿能力。

• 支持的语言: 中文、英文、日文、韩文、中文方言（粤语、四川话、上海话、天津话、武汉话等）
• 跨语言 & 混合语言: 支持零样本跨语言和代码切换场景的语音克隆。

实测效果很不错，4G显存就能跑。

克隆并安装

• 克隆仓库

  git clone --recursive https://github.com/FunAudioLLM/CosyVoice.git # If you failed to clone submodule due to network failures, please run following command until success cd CosyVoice git submodule update --init --recursive

• 安装 Conda: 请参阅 https://docs.conda.io/en/latest/miniconda.html

• 创建 Conda 环境：

  conda create -n cosyvoice python=3.10 conda activate cosyvoice # pynini is required by WeTextProcessing, use conda to install it as it can be executed on all platform. conda install -y -c conda-forge pynini==2.1.5 pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host=mirrors.aliyun.com # If you encounter sox compatibility issues # ubuntu sudo apt-get install sox libsox-dev # centos sudo yum install sox sox-devel

SDK模型下载 from modelscope import snapshot_download snapshot_download('iic/CosyVoice2-0.5B', local_dir='iic/CosyVoice2-0.5B')

import sys
sys.path.append('third_party/Matcha-TTS')
from cosyvoice.cli.cosyvoice import CosyVoice, CosyVoice2
from cosyvoice.utils.file_utils import load_wav
import torchaudio

cosyvoice = CosyVoice2('iic/CosyVoice2-0.5B', load_jit=False, load_trt=False, fp16=False)

# NOTE if you want to reproduce the results on https://funaudiollm.github.io/cosyvoice2, please add text_frontend=False during inference
# zero_shot usage
prompt_speech_16k = load_wav('zero_shot_prompt.wav', 16000)
for i, j in enumerate(cosyvoice.inference_zero_shot('收到好友从远方寄来的生日礼物，那份意外的惊喜与深深的祝福让我心中充满了甜蜜的快乐，笑容如花儿般绽放。', '希望你以后能够做的比我还好呦。', prompt_speech_16k, stream=False)):
    torchaudio.save('zero_shot_{}.wav'.format(i), j['tts_speech'], cosyvoice.sample_rate)

下面是设计了界面，方便使用这个功能

import sys
import os
import tkinter as tk
from tkinter import filedialog, messagebox, ttk
from tqdm import tqdm
import re
import threading
import torch
import torchaudio

sys.path.append('third_party/Matcha-TTS')

from cosyvoice.cli.cosyvoice import CosyVoice, CosyVoice2
from cosyvoice.utils.file_utils import load_wav

class TextToSpeechApp:
    def __init__(self, root):
        self.root = root
        self.root.title("文本转语音工具")
        self.root.geometry("800x600")
        self.root.resizable(True, True)
        
        # 模型变量
        self.cosyvoice = None
        self.prompt_speech_16k = None
        
        # 状态变量
        self.is_processing = False
        
        self.setup_ui()
        
    def setup_ui(self):
        """设置用户界面"""
        # 主框架
        main_frame = ttk.Frame(self.root, padding="10")
        main_frame.grid(row=0, column=0, sticky=(tk.W, tk.E, tk.N, tk.S))
        
        # 配置网格权重
        self.root.columnconfigure(0, weight=1)
        self.root.rowconfigure(0, weight=1)
        main_frame.columnconfigure(1, weight=1)
        
        # 标题
        title_label = ttk.Label(main_frame, text="文本转语音工具", font=("Arial", 16, "bold"))
        title_label.grid(row=0, column=0, columnspan=3, pady=(0, 20))
        
        # 文件选择区域
        file_frame = ttk.LabelFrame(main_frame, text="文件设置", padding="10")
        file_frame.grid(row=1, column=0, columnspan=3, sticky=(tk.W, tk.E), pady=(0, 10))
        file_frame.columnconfigure(1, weight=1)
        
        # 输入文件选择
        ttk.Label(file_frame, text="输入文本文件:").grid(row=0, column=0, sticky=tk.W, pady=5)
        self.input_file_var = tk.StringVar()
        ttk.Entry(file_frame, textvariable=self.input_file_var, state="readonly").grid(row=0, column=1, sticky=(tk.W, tk.E), padx=(5, 5), pady=5)
        ttk.Button(file_frame, text="浏览...", command=self.select_input_file).grid(row=0, column=2, pady=5)
        
        # 输出目录选择
        ttk.Label(file_frame, text="输出目录:").grid(row=1, column=0, sticky=tk.W, pady=5)
        self.output_dir_var = tk.StringVar()
        ttk.Entry(file_frame, textvariable=self.output_dir_var, state="readonly").grid(row=1, column=1, sticky=(tk.W, tk.E), padx=(5, 5), pady=5)
        ttk.Button(file_frame, text="浏览...", command=self.select_output_dir).grid(row=1, column=2, pady=5)
        
        # 设置区域
        settings_frame = ttk.LabelFrame(main_frame, text="转换设置", padding="10")
        settings_frame.grid(row=2, column=0, columnspan=3, sticky=(tk.W, tk.E), pady=(0, 10))
        settings_frame.columnconfigure(1, weight=1)
        
        # 示例音频文件选择
        ttk.Label(settings_frame, text="示例音频文件:").grid(row=0, column=0, sticky=tk.W, pady=5)
        self.prompt_file_var = tk.StringVar(value="./asset/zero_shot_prompt.wav")
        ttk.Entry(settings_frame, textvariable=self.prompt_file_var).grid(row=0, column=1, sticky=(tk.W, tk.E), padx=(5, 5), pady=5)
        ttk.Button(settings_frame, text="浏览...", command=self.select_prompt_file).grid(row=0, column=2, pady=5)
        
        # 示例文本
        ttk.Label(settings_frame, text="示例文本:").grid(row=1, column=0, sticky=tk.W, pady=5)
        self.prompt_text_var = tk.StringVar(value="希望你以后能够做的比我还好呦。")
        ttk.Entry(settings_frame, textvariable=self.prompt_text_var).grid(row=1, column=1, columnspan=2, sticky=(tk.W, tk.E), padx=(5, 0), pady=5)
        
        # 文本分段长度
        ttk.Label(settings_frame, text="分段长度:").grid(row=2, column=0, sticky=tk.W, pady=5)
        self.chunk_size_var = tk.StringVar(value="2000")
        chunk_size_frame = ttk.Frame(settings_frame)
        chunk_size_frame.grid(row=2, column=1, columnspan=2, sticky=(tk.W, tk.E), pady=5)
        ttk.Entry(chunk_size_frame, textvariable=self.chunk_size_var, width=10).grid(row=0, column=0, sticky=tk.W)
        ttk.Label(chunk_size_frame, text="字符").grid(row=0, column=1, sticky=tk.W, padx=(5, 0))
        
        # 文本预览区域
        preview_frame = ttk.LabelFrame(main_frame, text="文本预览", padding="10")
        preview_frame.grid(row=3, column=0, columnspan=3, sticky=(tk.W, tk.E, tk.N, tk.S), pady=(0, 10))
        preview_frame.columnconfigure(0, weight=1)
        preview_frame.rowconfigure(0, weight=1)
        main_frame.rowconfigure(3, weight=1)
        
        # 文本预览文本框
        self.text_preview = tk.Text(preview_frame, height=10, wrap=tk.WORD)
        text_scrollbar = ttk.Scrollbar(preview_frame, orient="vertical", command=self.text_preview.yview)
        self.text_preview.configure(yscrollcommand=text_scrollbar.set)
        self.text_preview.grid(row=0, column=0, sticky=(tk.W, tk.E, tk.N, tk.S))
        text_scrollbar.grid(row=0, column=1, sticky=(tk.N, tk.S))
        
        # 进度区域
        progress_frame = ttk.Frame(main_frame)
        progress_frame.grid(row=4, column=0, columnspan=3, sticky=(tk.W, tk.E), pady=(0, 10))
        progress_frame.columnconfigure(0, weight=1)
        
        self.progress_var = tk.DoubleVar()
        self.progress_bar = ttk.Progressbar(progress_frame, variable=self.progress_var, maximum=100)
        self.progress_bar.grid(row=0, column=0, columnspan=2, sticky=(tk.W, tk.E), pady=5)
        
        self.progress_label = ttk.Label(progress_frame, text="就绪")
        self.progress_label.grid(row=1, column=0, columnspan=2, sticky=tk.W)
        
        # 按钮区域
        button_frame = ttk.Frame(main_frame)
        button_frame.grid(row=5, column=0, columnspan=3, pady=10)
        
        self.start_button = ttk.Button(button_frame, text="开始转换", command=self.start_conversion)
        self.start_button.grid(row=0, column=0, padx=(0, 10))
        
        self.cancel_button = ttk.Button(button_frame, text="取消", command=self.cancel_conversion, state="disabled")
        self.cancel_button.grid(row=0, column=1)
        
        # 状态栏
        self.status_var = tk.StringVar(value="就绪")
        status_bar = ttk.Label(main_frame, textvariable=self.status_var, relief=tk.SUNKEN)
        status_bar.grid(row=6, column=0, columnspan=3, sticky=(tk.W, tk.E))
        
    def select_input_file(self):
        """选择输入文本文件"""
        file_path = filedialog.askopenfilename(
            title="选择文本文件",
            filetypes=[("文本文件", "*.txt"), ("所有文件", "*.*")]
        )
        if file_path:
            self.input_file_var.set(file_path)
            self.load_text_preview(file_path)
            
    def select_output_dir(self):
        """选择输出目录"""
        output_dir = filedialog.askdirectory(title="选择输出目录")
        if output_dir:
            self.output_dir_var.set(output_dir)
            
    def select_prompt_file(self):
        """选择示例音频文件"""
        file_path = filedialog.askopenfilename(
            title="选择示例音频文件",
            filetypes=[("音频文件", "*.wav"), ("所有文件", "*.*")]
        )
        if file_path:
            self.prompt_file_var.set(file_path)
            
    def load_text_preview(self, file_path):
        """加载文本预览"""
        try:
            with open(file_path, 'r', encoding='utf-8') as f:
                content = f.read()
                # 只显示前1000个字符作为预览
                preview_content = content[:1000] + ("..." if len(content) > 1000 else "")
                self.text_preview.delete(1.0, tk.END)
                self.text_preview.insert(1.0, preview_content)
                self.status_var.set(f"已加载文件: {os.path.basename(file_path)}")
        except Exception as e:
            messagebox.showerror("错误", f"读取文件失败: {str(e)}")
            
    def start_conversion(self):
        """开始转换"""
        if not self.input_file_var.get():
            messagebox.showwarning("警告", "请选择输入文本文件")
            return
            
        if not self.output_dir_var.get():
            messagebox.showwarning("警告", "请选择输出目录")
            return
            
        if not os.path.exists(self.prompt_file_var.get()):
            messagebox.showwarning("警告", "示例音频文件不存在")
            return
            
        # 禁用开始按钮，启用取消按钮
        self.start_button.config(state="disabled")
        self.cancel_button.config(state="normal")
        self.is_processing = True
        
        # 在新线程中运行转换过程
        thread = threading.Thread(target=self.run_conversion)
        thread.daemon = True
        thread.start()
        
    def cancel_conversion(self):
        """取消转换"""
        self.is_processing = False
        self.status_var.set("转换已取消")
        
    def run_conversion(self):
        """运行转换过程"""
        try:
            self.status_var.set("正在初始化模型...")
            self.progress_label.config(text="正在初始化模型...")
            
            # 初始化模型
            if not self.initialize_model():
                return
                
            # 处理文本文件
            self.process_text_file()
            
        except Exception as e:
            messagebox.showerror("错误", f"转换过程出错: {str(e)}")
        finally:
            # 恢复按钮状态
            self.root.after(0, self.reset_ui)
            
    def initialize_model(self):
        """初始化模型"""
        try:
            # 加载示例音频
            self.prompt_speech_16k = load_wav(self.prompt_file_var.get(), 16000)
            
            # 初始化模型
            self.cosyvoice = CosyVoice2(
                'pretrained_models/CosyVoice2-0.5B', 
                load_jit=False, 
                load_trt=False, 
                load_vllm=False, 
                fp16=False
            )
            return True
        except Exception as e:
            messagebox.showerror("错误", f"初始化模型失败: {str(e)}")
            return False
            
    def process_text_file(self):
        """处理文本文件"""
        file_path = self.input_file_var.get()
        output_dir = self.output_dir_var.get()
        
        # 创建输出目录
        if not os.path.exists(output_dir):
            os.makedirs(output_dir)
            
        # 读取文本文件
        try:
            with open(file_path, 'r', encoding='utf-8') as f:
                text = f.read()
        except Exception as e:
            messagebox.showerror("错误", f"读取文件失败: {str(e)}")
            return
            
        # 预处理文本
        text = re.sub(r'\s+', ' ', text).strip()
        
        # 分段处理
        chunk_size = int(self.chunk_size_var.get())
        chunks = self.split_text_into_chunks(text, chunk_size)
        
        total_chunks = len(chunks)
        self.root.after(0, lambda: self.progress_bar.config(maximum=total_chunks))
        
        # 处理每个文本块
        total_files = 0
        
        for i, chunk in enumerate(chunks):
            if not self.is_processing:
                break
                
            # 更新进度
            progress_text = f"正在转换第 {i+1}/{total_chunks} 段"
            self.root.after(0, lambda txt=progress_text: self.update_progress(i, txt))
            
            try:
                # 处理单个文本块，直接输出所有音频文件
                files_count = self.process_single_chunk(chunk, i+1)
                total_files += files_count
                    
            except Exception as e:
                error_msg = f"转换第 {i+1} 段时出错: {str(e)}"
                self.root.after(0, lambda msg=error_msg: self.status_var.set(msg))
                continue
        
        if self.is_processing:
            self.root.after(0, lambda: messagebox.showinfo("完成", 
                f"转换完成！共生成 {total_files} 个音频文件。\n输出目录: {output_dir}"))
            
    def split_text_into_chunks(self, text, chunk_size):
        """将文本分割成块"""
        if len(text) <= chunk_size:
            return [text]
            
        sentences = re.split(r'[。！？!?]', text)
        sentences = [s.strip() for s in sentences if s.strip()]
        
        chunks = []
        current_chunk = ""
        
        for sentence in sentences:
            if len(current_chunk) + len(sentence) + 1 <= chunk_size or not current_chunk:
                if current_chunk:
                    current_chunk += "。" + sentence
                else:
                    current_chunk = sentence
            else:
                chunks.append(current_chunk + "。")
                current_chunk = sentence
        
        if current_chunk:
            chunks.append(current_chunk + "。")
            
        return chunks
        
    def process_single_chunk(self, chunk, chunk_index):
        """处理单个文本块，直接输出所有音频文件"""
        files_count = 0
        
        try:
            # 使用零样本推理，直接保存每个生成的音频片段
            for j, result in enumerate(self.cosyvoice.inference_zero_shot(
                chunk, 
                self.prompt_text_var.get(), 
                self.prompt_speech_16k, 
                stream=False
            )):
                # 为每个片段生成单独的文件
                output_filename = os.path.join(self.output_dir_var.get(), f'audio_{chunk_index:03d}_{j+1:03d}.wav')
                torchaudio.save(output_filename, result['tts_speech'], self.cosyvoice.sample_rate)
                files_count += 1
                
        except Exception as e:
            raise Exception(f"处理文本块时出错: {str(e)}")
            
        return files_count
            
    def update_progress(self, value, text):
        """更新进度条"""
        self.progress_var.set(value)
        self.progress_label.config(text=text)
        self.status_var.set(text)
        
    def reset_ui(self):
        """重置UI状态"""
        self.start_button.config(state="normal")
        self.cancel_button.config(state="disabled")
        self.progress_var.set(0)
        self.progress_label.config(text="就绪")
        self.is_processing = False

def main():
    root = tk.Tk()
    app = TextToSpeechApp(root)
    root.mainloop()

if __name__ == "__main__":
    main()