Whisper Assistant is an extension for Visual Studio Code that transcribes your spoken words into text within the VSCode editor. This hands-free approach to coding allows you to focus on your ideas instead of your typing.
Whisper Assistant can also be integrated with other powerful AI tools, such as Chat GPT-4 or the Cursor.so application, to create a dynamic, AI-driven development environment.
Whisper Assistant utilizes the Whisper AI locally, offering a free voice transcription service.
By default, the base model of Whisper AI is used, balancing accuracy and performance. You can select a different model in the extension settings, but remember to download your chosen model before using Whisper Assistant. Failure to download the selected model can lead to errors. The base model is recommended and set as default.
For more details about Whisper, visit the Whisper OpenAI GitHub page.
To install and setup Whisper Assistant, follow these steps:
-
Install SoX to enable easy microphone recording through the command line.
- MacOS: Using the Homebrew package manager, run the following command in your terminal:
brew install sox
- Windows: Using the Chocolatey package manager, run the following command in your terminal:
choco install sox.portable
- Ubuntu: Run the following command in your terminal:
sudo apt install sox
- MacOS: Using the Homebrew package manager, run the following command in your terminal:
-
Install Whisper locally. This requires the following prerequisites:
- Install Python 3
- Install PIP
- Follow the instructions on the Whisper OpenAI GitHub page to complete the Whisper installation.
-
Install the Whisper Assistant extension into Visual Studio Code or the Cursor.so application.
- Initialization: Upon loading Visual Studio Code, the extension verifies the correct installation of SoX and Whisper. If any issues are detected, an error message will be displayed. These dependencies must be installed to use the extension.
Once initialization is complete, a quote icon will appear in the bottom right status bar.
- Starting the Recording: Activate the extension by clicking on the quote icon or using the shortcut
Command+M
(for Mac) orControl+M
(for Windows). You can record for as long as you like, but remember, the longer the recording, the longer the transcription process. The recording time will be displayed in the status bar.
- Stopping the Recording: Stop the recording using the same shortcut (
Command+M
orControl+M
). The extension icon in the status bar will change to a loading icon, and a progress message will be displayed, indicating that the transcription is underway.
- Transcription: Once the transcription is complete, the text will be saved to the clipboard. This allows you to use the transcription in any program, not just within Visual Studio Code. If an editor is active, the transcription will be pasted there automatically.
Tip: A good microphone will improve transcription accuracy, although it is not a requirement.
Tip: For an optimal experience, consider using the Cursor.so application to directly call the Chat GPT-4 API for code instructions. This allows you to use your voice to instruct GPT to refactor your code, write unit tests, and implement various improvements.
To enhance your development experience with Cursor.so and Whisper Assistant, follow these simple steps:
- Start the recording: Press
Command+M
(Mac) orControl+M
(Windows). - Speak your instructions clearly.
- Stop the recording: Press
Command+M
(Mac) orControl+M
(Windows). Note: This initiates the transcription process. - Open the Cursor dialog: Press
Command+K
orCommand+L
. Important: Do this before the transcription completes. - The transcribed text will automatically populate the Cursor dialog. Here, you can edit the text or add files/docs, then press
Enter
to execute the GPT query.
By integrating Cursor.so with Whisper Assistant, you can provide extensive instructions without the need for typing, significantly enhancing your development workflow.
Please note that this extension has been primarily tested on Mac OS. While efforts have been made to ensure compatibility, its functionality on other platforms such as Windows or Linux cannot be fully guaranteed. I welcome and appreciate any pull requests to address potential issues encountered on these platforms.