Open source macOS dictation app that turns speech into text with on-device models and optional AI cleanup, a local alternative to Wispr Flow.
akshayaggarwal99/jarvis-ai-assistant
Open-source voice dictation app for Mac: hold Fn, speak, and cleaned-up text is typed wherever your cursor is.
Open-source alternative to:Wispr Flow
- Stars
- 659
- Forks
- 114
- Open issues
- 7
- Last commit
- 29 days ago
What is it?
Jarvis is a macOS voice assistant centered on dictation. You hold the Fn key, speak, and release it; the app transcribes the audio, removes filler words, fixes grammar, and inserts the resulting text into whatever app has focus. It can also rephrase or reformat text and run small actions such as opening a website or setting a timer.
Transcription can run fully offline using local Whisper models or NVIDIA Parakeet through Sherpa-ONNX. For post-processing, it supports local LLMs through Ollama, or a cloud setup using Deepgram for speech-to-text and Gemini for text cleanup. Prompts that control formatting and assistant behavior can be customized.
The app is distributed as a signed and notarized DMG for Apple Silicon and Intel Macs, with an iOS version available through TestFlight. It is built with TypeScript and Node.js, and the source can be built locally with npm. A Windows version is listed on the roadmap but not yet available.
Key features
- 01Hold Fn to record, release to insert cleaned-up text into any app.
- 02Removes filler words and fixes grammar automatically.
- 03Offline transcription with local Whisper or NVIDIA Parakeet models.
- 04Local LLM post-processing through Ollama.
- 05Cloud option using Deepgram and Gemini.
- 06Customizable prompts for formatting and assistant behavior.
Use cases
- Dictating emails, messages, or documents faster than typing on a Mac.
- Keeping voice data on-device by running Whisper or Parakeet locally.
- Turning rough spoken notes into punctuated, well-formatted text.
- Using a local Ollama model for private text rewriting.
Quick start
- 1
Download the DMG from the releases page and open the app
- 2
Clone the repository
git clone https://github.com/akshayaggarwal99/jarvis-ai-assistant.git - 3
Enter the project folder and install, build and run from source
cd jarvis-ai-assistant && npm install && npm run build && npm run dev
Things to know
- Currently available only for macOS; the Windows version is still on the roadmap.
- Build from source requires Node.js 18+, macOS 10.13+ and Xcode CLI tools.
- Fully offline use requires local models, and AI post-processing via Ollama needs Ollama installed separately.
- Cloud mode needs your own Deepgram and Gemini API keys.
Topics
FAQ
- What is jarvis-ai-assistant?
- Open-source voice dictation app for Mac: hold Fn, speak, and cleaned-up text is typed wherever your cursor is. Jarvis is a macOS voice assistant centered on dictation. You hold the Fn key, speak, and release it; the app transcribes the audio, removes filler words, fixes grammar, and inserts the resulting text into whatever app has focus. It can also rephrase or reformat text and run small actions such as opening a website or setting a timer.
- Is jarvis-ai-assistant free to use? What license does it have?
- jarvis-ai-assistant is open source under the MIT license. Check the license terms before using it in commercial products.
- Is jarvis-ai-assistant still maintained?
- Actively maintained. The last commit to jarvis-ai-assistant was on Sep 23, 2026.
- What can I use jarvis-ai-assistant for?
- Dictating emails, messages, or documents faster than typing on a Mac. Keeping voice data on-device by running Whisper or Parakeet locally. Turning rough spoken notes into punctuated, well-formatted text. Using a local Ollama model for private text rewriting.
- How do I install jarvis-ai-assistant?
- Download the DMG from the releases page and open the app Clone the repository: git clone https://github.com/akshayaggarwal99/jarvis-ai-assistant.git Enter the project folder and install, build and run from source: cd jarvis-ai-assistant && npm install && npm run build && npm run dev
- Where can I get jarvis-ai-assistant?
- The source code and installation instructions are on GitHub: https://github.com/akshayaggarwal99/jarvis-ai-assistant
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Native macOS speech-to-text app that transcribes dictation on-device or via cloud APIs, then reprocesses text with reusable AI workflows.
Open-source, local-first AI dictation app that turns speech into text on macOS and Windows, working offline with Whisper.
Local-first AI subtitle generator that runs offline and plugs into DaVinci Resolve, Premiere Pro and After Effects.
Privacy-first AI meeting assistant that records, transcribes and summarizes meetings locally on macOS and Windows, with optional Ollama or cloud LLM summaries.
Desktop app that transcribes audio and video fully offline on your own device, using Whisper and other speech models.