python实践-实现实时语音转文字本地部署版（二）

news2025/4/19 11:27:03

一、技术栈

python 3.10.6

vosk 需下载对应模型（vosk-model-cn-0.22）模型下载慢的同学看最后的资源链接。

pyaudio

keyboard

二、实现功能

本地化实现麦克风语音录入，实时生成文字，并保存至本地文档。

三、实现代码

from vosk import Model, KaldiRecognizer
import pyaudio  # 用于实时音频输入
import json  # 用于解析识别结果
import keyboard
import threading
import sys  # 引入 sys 模块用于退出程序

# 指向模型根目录
model_path = "D:/2.DevEnv/vosk/vosk-model-cn-0.22"  # 注意路径使用正斜杠或双反斜杠
model = Model(model_path)
rec = KaldiRecognizer(model, 16000)  # 采样率设为 16000，与模型匹配

# 初始化 PyAudio
p = pyaudio.PyAudio()

# 打开音频流
stream = p.open(format=pyaudio.paInt16, channels=1, rate=16000, input=True, frames_per_buffer=4000)
stream.start_stream()

recording = False
lock = threading.Lock()

# 定义结束关键词
END_KEYWORDS = ["结束", "停止"]
# 用于存储识别出来的文本
recognized_text = ""

def start_recognition():
    global recording, recognized_text
    with lock:
        recording = True
    print("开始录音，请说话...")
    try:
        while True:
            with lock:
                if not recording:
                    break
            data = stream.read(4000)
            if rec.AcceptWaveform(data):
                result = json.loads(rec.Result())
                text = result.get("text", "")
                print(text)
                # 累加识别的文本
                recognized_text += text + " "
                # 检查是否包含结束关键词
                for keyword in END_KEYWORDS:
                    if keyword in text:
                        stop_recognition()
                        break
    except Exception as e:
        print(f"录音过程中出现错误: {e}")

def stop_recognition():
    global recording, recognized_text
    with lock:
        if not recording:
            return
        recording = False
    print("停止录音")
    try:
        final_result = json.loads(rec.FinalResult())
        final_text = final_result.get("text", "")
        print("Final result:", final_text)
        # 累加最终识别的文本
        recognized_text += final_text
    except Exception as e:
        print(f"获取最终结果时出现错误: {e}")

    # 关闭音频流和 PyAudio
    try:
        stream.stop_stream()
        stream.close()
        p.terminate()
    except Exception as e:
        print(f"关闭音频流时出现错误: {e}")

    # 保存识别的文本到文件
    try:
        with open('recognized_text.txt', 'w', encoding='utf-8') as file:
            file.write(recognized_text)
        print("识别的文本已保存到 recognized_text.txt")
    except Exception as e:
        print(f"保存文件时出现错误: {e}")

    # 终止程序运行
    sys.exit(0)

def on_space_press(event):
    global recording
    if event.name == 'space':
        with lock:
            if not recording:
                thread = threading.Thread(target=start_recognition)
                thread.start()
            else:
                stop_recognition()

# 监听空格键事件
keyboard.on_press(on_space_press)

print("按空格键开始识别，说 '结束' 或 '停止' 来结束录音。")
keyboard.wait()