OpenAI Whisper Speech-to-Text Transcriber

Overview

This script records audio using a hotkey, saves it as a WAV file, and then transcribes it into text using the OpenAI Whisper model. The transcribed text is automatically typed where your cursor is.

Screenshots

Requirements

• Python 3.x
• pyaudio
• wave
• threading
• keyboard
• whisper
• pyautogui
• argparse

You can install these packages using pip:

pip install -r requirements.txt

ffmpeg

This program requires FFMPEG. The easiest way to install it on Windows is using Chocolatey, a package manager. Here’s how:

• install Chocolatey by following the instructions on the official website.
• Once Chocolatey is installed, open a command prompt or PowerShell as an administrator.
• Run the following command to install FFMPEG:

choco install ffmpeg

Now, FFMPEG will be installed and automatically added to your system's PATH.

Usage

Basic Usage

To run the script, use the following command:

python recorder.py

By default the hotkey to start and end the recording is "alt+x". The program will read in the audio from the first hotkey press to the second hotkey press. After the audio is saved and transcribed, the transcribed audio will be typed out wherever the cursor is currently placed. No need to keep the program shell front and center, the program will continue to monitor the hotkey in the background.

Custom Hotkey

You can specify a custom hotkey for toggling recording:

python recorder.py --hotkey '<key-1>+<key-2>'

Custom language

You can specify a custom language for transcribing:

python recorder.py --language '<language>'

Custom model

You can specify a model for transcribing:

python recorder.py --model '<model-size>'

Console Language

You can specify a language for the console output:

python recorder.py --console '<language>'

Now supported languages are:

• English (en)
• Chinese (zh)
• Traditional Chinese (zh-tw)

Run on startup

1. Open the run dialog by pressing Win+R.
2. Type shell:startup and press enter.
3. Create a shortcut to run.bat (or 執行.bat) in the folder that opens.

Class AudioRecorder

Methods

• toggle_recording(): Toggles the recording state.
• record(): Records audio while self.recording is True.
• transcribe_recording(): Transcribes the recorded audio.
• save_audio(): Saves the recorded audio and transcribes it.
• set_hotkey(hotkey): Sets a hotkey for toggling recording.

License

MIT