๐๏ธ OpenSource Voice Dictation Agent (Wispr Flow clone)
Jarvis AI Assistant is an open-source voice dictation desktop application that replicates the functionality of commercial tools like Wispr Flow. It enables users to dictate text directly into any application on their computer with minimal latency. Built with Electron, it offers cross-platform support for macOS, Windows, and Linux. The application emphasizes privacy and offline capabilities, processing speech locally whenever possible. The tool supports multiple speech recognition engines: Whisper models (via whisper.cpp) for multilingual transcription and NVIDIA's Parakeet TDT model (via Sherpa-ONNX) for fast, on-device processing. It also provides cloud fallback to Google Gemini for enhanced transcription or AI-powered actions. Users can choose between engines based on their needs for speed, accuracy, or language support. Key features include hands-free mode for uninterrupted dictation, a macOS iOS keyboard extension for mobile use, and seamless integration with system text fields. The application optimizes audio capture to prevent zero-byte recordings and includes fixes for race conditions, ensuring stable long-duration dictation sessions. As an open-source project, Jarvis AI Assistant welcomes community contributions and is actively maintained with regular updates. It aims to democratize access to high-quality voice dictation by removing paywalls and reducing reliance on proprietary cloud services.
- Category
- AI Agents
- Pricing
- Open Source
Problem solved
Many users require fast and accurate voice dictation but are limited by paid, cloud-dependent solutions that may compromise privacy. Jarvis AI Assistant addresses this by offering a free, open-source dictation tool that works offline, giving users full control over their data without sacrificing features.
Use cases
- Dictating emails, documents, and chat messages hands-free.
- Enabling accessibility for users with motor impairments through voice-controlled computing.
- Multilingual dictation using local Whisper or Parakeet models.
- Using voice commands to trigger AI-powered actions via Gemini integration.
- Offline dictation in sensitive or low-connectivity environments.
Audience
- Developers
- Business Users
- Content Creators
- Hobbyists
Pros
- Completely free and open source.
- Supports local offline speech recognition with multiple model backends (Whisper, Parakeet).
- System-wide dictation across all desktop applications.
- Hands-free mode for continuous dictation without manual activation.
- Active development with frequent updates and community engagement.
Cons
- Requires technical setup including model downloads and native dependencies.
- May have platform-specific stability issues (e.g., macOS audio race conditions).
- Performance and accuracy may lag behind commercial alternatives under certain conditions.
Official website: https://github.com/akshayaggarwal99/jarvis-ai-assistant
Canonical LambdaBase page: https://www.lambdabase.com/tools/opensource-voice-dictation-agent-wispr-flow-clone