Seamless dictation powered by Groq Whisper
This is an MOD of the outstanding Whispering App (not from me) and and I don't change much. The only change is that it's to use the Groq API, not the OpenAI API. By the way: "Groq or Grok : Groq is not Grok. Grok is the AI model developed by X/Twitter whereas Groq is a new technology for chips allowing for efficient LLM execution." That means I only changed four lines in this file: packages/shared/src/services/TranscriptionServiceWhisperingLive.ts
And I would recommend to use the original version, if he maybe published a version with Groq support, in the future. Tipp: After the installation, you have to re-register the shortcuts, if you had an OpenAI version before. Open Question: I'm not sure if a auto update function is still implemented, but then it will override this version with a standard version.
You get your faster and more precise Whisper 3 API model from Groq here: https://console.groq.com/keys
Groq supports the newer whisper V3 model via API (OpenAI supports only V2 via API). Groq is 10 times cheaper. It has less delay: https://wow.groq.com/artificialanalysis-ai-llm-benchmark-doubles-axis-to-fit-new-groq-lpu-inference-engine-performance-results/ In my experience, it is more precise expressly with shorter sentences and in general more accurate. Quality comparison https://github.com/openai/whisper
Whispering Fork is an open-source transcription application that provides global speech-to-text functionality, with options such as keyboard shortcuts and automatic copy and paste to make dictating as seamless as possible. Here's the original project for the OpenAI API. https://github.com/braden-w/whispering From here on, it's a copy of the original text...
Under the hood, it's powered by Groq's Whisper API, making it significantly more accurate than built-in dictation.
Desktop.mp4
Extension.mp4
- Global Transcription: Access Whisper's speech-to-text functionality anywhere with a global keyboard shortcut or within two button clicks.
- Cross-Platform Experience:
- Desktop App: Enables global transcription across all applications.
- Browser Extension: Provides global transcription in the browser by communicating with the web app.
- Chat Integration: The browser extension adds a recording button to ChatGPT and Claude websites, allowing direct voice input and transcription in the chat interface.
- Transcription Management: Review and edit transcriptions within the Whispering app to ensure accuracy and clarity.
- Automatic Clipboard Integration: Once transcription is complete, the text is automatically copied to your clipboard. An option for automatic pasting is also available.
Visit whispering.bradenwong.com, which has the latest version of the apps/app
folder hosted on Vercel.
Install the Chrome Extension from the Chrome Web Store here.
To download and install the Whispering desktop app, follow the steps for your operating system:
- Download the installation package ending in
.msi
from the latest releases page. - Open the downloaded
.msi
file to run the installer. - If prompted as unsafe, click on
More Info
->Run Anyway
to proceed with the installation. - Follow the on-screen instructions to complete the installation.
For macOS, follow these steps to install Whispering:
-
Download the installation package:
- Visit the latest releases page.
- Choose the appropriate package:
- For Apple Silicon:
Whispering_x.x.x_aarch64.dmg
- For Intel:
Whispering_x.x.x_x64.dmg
- For Apple Silicon:
-
Install the application:
- Open the downloaded
.dmg
file. - Drag the Whispering app to the Applications folder.
- Open the downloaded
-
Launch Whispering:
- Open Whispering from the Applications folder.
- If you see a warning about unverified developer:
- Click
Cancel
- Right-click the app in Finder and select
Open
- Click
-
Troubleshooting (Apple Silicon only): If you encounter the error
"Whispering" is damaged and can't be opened
:- Open Terminal
- Run the following command:
xattr -cr /Applications/Whispering.app
- Try opening the application again
After completing these steps, Whispering should be ready to use on your macOS system.
For Linux, there are multiple installation options available. Choose the one that suits your environment:
-
AppImage:
- Download the
.AppImage
file from the latest releases page. - Make the file executable:
chmod +x whispering_x.x.x_amd64.AppImage
- Run the AppImage:
./whispering_x.x.x_amd64.AppImage
- Download the
-
DEB Package (Debian/Ubuntu):
- Download the
.deb
file from the latest releases page. - Install the package using
dpkg
:sudo dpkg -i whispering_x.x.x_amd64.deb
- Resolve any dependency issues:
sudo apt-get install -f
- Download the
After installation, the app will be ready to use.
After installing the Chrome Extension, you will find a Whispering icon on the Chrome extensions bar. Click on this icon to open the extension. Click the microphone button to record your voice, and then click the square button when you're done. Your transcription will appear in the text box below.
To access the ChatGPT or Claude feature, navigate to the ChatGPT or Claude web page. You'll see a new recording button in the chat interface. Click this button to start and stop recording, and the transcribed text will be automatically inserted into the chat input field.
To access the shortcut feature, press Control + Shift + X or Command + Shift + X (later configurable through Chrome extension shortcuts) to start recording from any website. The transcription will be automatically copied into your clipboard and paste into the current input field.
The Chrome Extension communicates with whispering.bradenwong.com, and will automatically attempt to create a tab in the background if it does not find one. Most bugs can be attributed to this communication failing, such as rare instances where the tab falls asleep.
The web app is accessible via whispering.bradenwong.com. Click the microphone button to record your voice, and then click the square button when you're done. Your transcription will appear in the text box.
After installing the Whispering desktop app, press Control/Command + Shift + ; (configurable in settings) to start recording from anywhere on your desktop. The transcription will be automatically copied into your clipboard and pasted, though both features can be toggled in the settings.
The Whispering app is built using the following technologies and libraries:
- Svelte 5: The UI reactivity library of choice.
- SvelteKit: For routing and static site generation, used for making both the website and the static frontend for the Tauri app.
- Tauri: The desktop app framework.
- Effect-TS: To sprinkle some functional programming and write extremely type-safe functions, where errors are included in the return type of the function signature.
- Svelte Sonner: A simple, customizable toast library for Svelte applications. Used to capture and display errors bubbled up via Effect-TS using the
renderAsToast
function. - TanStack Table: To power all data tables.
- IndexedDB: A low-level API for storing large amounts of structured data in the browser. Synchronized with Tanstack Table using an Effect service in RecordingDbServiceIndexedDbLive.svelte.ts.
- ShadCN-Svelte: The UI component library of choice.
- TailwindCSS: A utility-first CSS framework for rapidly building custom user interfaces.
- Turborepo: For monorepo management, so that
apps/app
andapps/extension
can share the same codebase, drastically reducing code duplication and more importantly, keeping a single source of truth. - Rust: For extending desktop app features, such as using the
enigo
crate for handling automatic pasting. - Vercel: Hosting that's decent for a hobby project and has nice Turborepo integrations.
- Zapsplat.com: A royalty-free sound effects library.
The Whispering Chrome extension is built using:
- Plasmo: A framework for building Chrome extensions. We use the relay flow to communicate to the Whispering website.
- Effect-TS: To sprinkle some functional programming and write extremely type-safe functions, where errors are included in the return type of the function signature.
- React: The UI reactivity library for the Chrome extension, as Plasmo unfortunately doesn't support Svelte 5.
- ShadCN: The UI component library for the Chrome extension.
- TailwindCSS: A utility-first CSS framework for rapidly building custom user interfaces.
- Chrome API: The Chrome extension API.
- Zapsplat.com: A royalty-free sound effects library.
To set up the project on your local machine, follow these steps:
- Clone the repository:
git clone https://github.com/braden-w/whispering.git
- Change into the project directory:
cd whispering
- Install the necessary dependencies:
pnpm i
To run the Whispering desktop app and website in development mode:
- Change into the app directory:
cd apps/app
- Run the development server:
pnpm tauri dev
The desktop app should automatically open for local development. To develop the web app, open your browser and navigate to http://localhost:5173
.
To run the Whispering Chrome extension in development mode:
- Change into the extension directory:
cd apps/extension
- Run the development server:
pnpm dev --target=chrome-mv3
To develop the chrome extension, load it into Chrome by navigating to chrome://extensions
, enabling developer mode, and loading the apps/extension/build/{platform}-{manifest-version}-dev
folder as an unpacked extension.
If you ever have concerns regarding the trustworthiness of the installers or would like more control, you can always build the executable yourself. This requires more setup, but it ensures that you are running the code you expect. Such is the beauty of open-source software!
- Change into the extension directory:
cd apps/extension
- Install the necessary dependencies:
pnpm i
- Run Plasmo build:
pnpm plasmo build --target=chrome-mv3
- Output should be found in
apps/extension/build/chrome-mv3-prod
, which can be loaded into Chrome as an unpacked extension. - Alternatively, you can build the extension for the Chrome Web Store:
pnpm plasmo build --target=chrome-mv3 --release
- Change into the extension directory:
cd apps/extension
- Install the necessary dependencies:
pnpm i
- Run Plasmo build:
pnpm plasmo build --target=firefox-mv3
- Output should be found in
apps/extension/build/firefox-mv3-prod
, which can be loaded into Chrome as an unpacked extension. - Alternatively, you can build the extension for the Chrome Web Store:
pnpm plasmo build --target=firefox-mv3 --release
- Change into the app directory:
cd apps/app
- Install the necessary dependencies:
pnpm i
- Run Tauri Build:
pnpm tauri build
- You can find the executable in the
apps/app/target/release
directory.
We welcome contributions from the community! If you'd like to contribute to Whispering, please follow these steps:
- Fork the repository.
- Create a new branch for your feature or bugfix:
git checkout -b feature/your-feature-name
orgit checkout -b fix/your-bugfix-name
- Make your changes and commit them with a descriptive message.
- Push your branch to your forked repository:
git push origin your-branch-name
- Create a pull request from your forked repository to the original one.
Please ensure your code follows established conventions and is well-documented.
Whispering is released under the MIT License.
This project is supported by the following amazing people and organizations:
If you encounter any issues or have suggestions for improvements, please open an issue on the GitHub issues tab or contact me via [email protected]. I really appreciate your feedback!
Thank you for using Whispering and happy writing!