speech_recognitionを使ってマイクに喋りかけて音声をwhisperで文字起こしする

pip install SpeechRecognition
pip install -U openai-whisper

import tkinter as tk
from tkinter import scrolledtext as tk_scrolledtext
import time
from tkinter import messagebox
import os

import threading
import queue
import speech_recognition as sr
import pyaudio

SAMPLE_RATE = 44100 
model_name='medium'
# ハルシネーションで出力される可能性があるテキストを定義
hallucinationTexts = [
        "ご視聴ありがとうございました",
        "Thanks for watching!",
]


# 録音スレッドの処理
def record_audio():
    global recording 
    global th2 
    global q    

    while recording:  # recordingがTrueの間、録音を継続
        print('record_audio')

        with sr.Microphone(sample_rate=SAMPLE_RATE) as source:
            r.dynamic_energy_threshold = True  # type: bool
            r.adjust_for_ambient_noise(source, duration=0.2)
            print("[なにか話してください]")
            audio = r.listen(source)
#
        print("音声処理中 ...")
#
        q.put(audio)
        print(str(q.qsize())+'Queue')
        if not th2.is_alive():
            th2 = threading.Thread(target=audio_to_text)
            th2.start()

"""
"""

# 音声を文字に変換する関数
def audio_to_text():
    global q    

    print('audio_to_text')

    while not q.empty():
        print('audio_to_text2')
        print(str(q.qsize())+'Queue')

        audio2 = q.get()

        try:
            text = r.recognize_google(audio2, language="ja")
        except sr.UnknownValueError:
            text = "Google Speech Recognition could not understand audio"
        except sr.RequestError as e:
            text = "Could not request results from Google Speech Recognition service; {0}".format(e)
        print('Google Speech :')

        print( f'{text}\n')

        try:
            text = r.recognize_whisper(audio2, model_name, language="ja", translate=False)

        except sr.UnknownValueError:
            text = "whisper Speech Recognition could not understand audio"
        except sr.RequestError as e:
            text = "Could not request results from whisper Speech Recognition service; {0}".format(e)
        print('whisper :')

        print( f'{text}\n')

        if text in hallucinationTexts:
            continue
        else :

            output_text.insert(tk.END, f'{text}\n')
            output_text.see(tk.END)
        time.sleep(0.1)

# 
def start_recording():
    global recording
    global th
    global th2 
    recording = True
    th = threading.Thread(target=record_audio)
    th.start()
    print("start_recording ...")

    th2 = threading.Thread(target=audio_to_text)
    th2.start()
# 録音を停止する関数
def stop_recording():
    global recording
    recording = False
#    th.join()
    print("stop_recording ...")

# 録音ボタンがクリックされたときの処理
def toggle_recording():
    global recording
    if recording:
        # 録音を停止
        stop_recording()
        record_button.config(text="Start Recording")
    else:
        # 録音を開始
        start_recording()
        record_button.config(text="Stop Recording")

def delete_window():
    global recording
    global starting

    global th
    global th2 
    global q    
    # 録音を停止
    stop_recording()
    record_button.config(text="Start Recording")
    record_button["state"] = "disabled"
  
    if   not starting :
        root.destroy()
    elif q.empty() and  not th.is_alive() and not th2.is_alive():
        print(str(q.qsize())+'Queue')
        root.destroy()
    else:
        messagebox.showerror('error', str(q.qsize())+'Queue or running. Wait a minute')


if __name__ == "__main__":
# オーディオキューを作成
    q = queue.Queue()

# 録音中かどうかを示すフラグ
    recording = False
    starting = False

# Tkinterウィンドウを作成
    root = tk.Tk()
    root.title("Speech Recognition App")

# 録音ボタンを作成
    record_button = tk.Button(root, text="Start Recording", command=toggle_recording) 
    record_button.pack(fill="x")

# 認識結果を表示するテキストウィジェットを作成
    output_text = tk_scrolledtext.ScrolledText(root, width=70, height=10, font=('Arial', 14))
    output_text.pack(pady=10)
    root.protocol("WM_DELETE_WINDOW", delete_window)

    print('Loading model...')
    global r
    r = sr.Recognizer()
    with sr.Microphone(sample_rate=SAMPLE_RATE) as source:
  #  
            r.dynamic_energy_threshold = True  # type: bool

            r.adjust_for_ambient_noise(source, duration=0.2)
            audio = r.listen(source)
    text = r.recognize_whisper(audio, model_name, language="ja", translate=False)
    print('Done')
    # GUIを表示
    root.mainloop()
    

@echo off
call %~dp0\scripts\env_for_icons.bat  %*
SET PATH=%PATH%;%WINPYDIRBASE%\PortableGit;%WINPYDIRBASE%\PortableGit\bin
SET PATH=%PATH%;%WINPYDIRBASE%\ffmpeg\bin
If not exist %WINPYDIRBASE%\content mkdir  %WINPYDIRBASE%\content 

set APP_NAME=openai-whisper
set APP_DIR=%WINPYDIRBASE%\content\%APP_NAME%
echo %APP_DIR%
md %APP_DIR%
cd %APP_DIR%
if not defined VENV_DIR (set "VENV_DIR=%APP_DIR%\venv")
if EXIST %VENV_DIR% goto :activate_venv


::python.exe -m venv "%VENV_DIR%" 
python.exe -m venv "%VENV_DIR%" --system-site-packages 
if %ERRORLEVEL% == 0 goto :activate_venv
echo Unable to create venv 
goto :skip_venv

:activate_venv
call "%VENV_DIR%\Scripts\activate"
::pip install -r requirements.txt
::If  exist %WINPYDIRBASE%\content\%APP_NAME%\ goto :skip_cmd
"D:\WinPython\Spyder.exe"
cmd.exe /k
goto :skip_venv
:skip_cmd
::python.exe webui.py
:skip_venv

timeout /t 55
::
cmd.exe /k