This project enables speech-to-text functionality on Linux using OpenAI's Whisper. It allows users to record audio with a press of a key (F9 by default), transcribe the audio to text using Whisper, and automatically paste the transcribed text where the cursor is located. This functionality is particularly useful for quickly converting speech into text without navigating through multiple applications.
Before you start, ensure you have the following installed on your Linux system:
ffmpeg for audio recording.
xclip for clipboard management.
xdotool for automated paste.
whisper for audio transcription.
A Python virtual environment with Whisper installed ($HOME/env_sandbox in this guide, if yours is different, modify stop_and_process_recording.sh with yours).
You can install ffmpeg xdotool and xclip using your distribution's package manager. For Whisper, follow the installation instructions provided by OpenAI.
Clone this repository to your local machine's home:
cd
git clone https://github.com/franchesoni/s2t.git
cd linux-speech-to-text
Disable Key Repeat for F9: Add xset -r 75 to your ~/.profile to disable key repeat for F9. Log out and log back in for the change to take effect.
Set Up xbindkeys: Install xbindkeys and create a .xbindkeysrc in your home directory with the following content:
"/home/user/s2t/start_recording.sh"
m:0x0 + c:75
"/home/user/s2t/stop_and_process_recording.sh"
Release + m:0x0 + c:75
Replace /home/user/s2t/ with the actual path to the cloned repository (your home!).
Modify the Scripts: Ensure start_recording.sh and stop_and_process_recording.sh are executable. If necessary, modify the scripts to match your system configuration, such as the path to your Python virtual environment.
chmod +x start_recording.sh stop_and_process_recording.sh
Open two consoles and cd to the project. Run ./start_recording.sh in one and say something. Run stop_and_process_recording.sh on the second. You should see that the model is being downloaded and that after it is, you have the transcript pasted where you had the cursor (it's also available using ctrl+v).
Start Recording: Press and hold F9 to start recording your speech. Stop Recording and Transcribe: Release F9 to stop recording. The audio will be transcribed to text using Whisper, and the text will be automatically pasted from the clipboard to where your cursor is located.
If the transcription does not work, ensure Whisper is correctly installed in your virtual environment and that the scripts point to the correct path of the virtual environment.
Make sure ffmpeg and xclip are correctly installed and accessible from your PATH.
Contributions are welcome! If you have improvements or bug fixes, please open a pull request or issue.
https://chat.openai.com/share/db168b73-7dfe-4467-a450-35791e48403b