AI · GitHub repo
omi
Open-source AI wearable and software that captures conversations and screen context.
Omi (originally started as Friend) is an open-source AI platform combining wearable hardware with desktop and mobile apps. It records audio and screen context in real time to generate transcriptions, notes, and action items, backed by an AI chat with long-term memory. Users can buy pre-built devices or build their own hardware using off-the-shelf microcontrollers.
What it's for
- Transcribe daily meetings and in-person conversations continuously
- Query an AI assistant about things you previously discussed or viewed on screen
- Extract automated summaries and action items from conversations
- Build custom AI wearable hardware using open schematics and firmware
Features
- Continuous real-time audio transcription with speaker diarization
- Screen capture integration on desktop platforms
- Memory-augmented AI chat interface to recall past conversations
- Companion mobile apps for iOS and Android built with Flutter
- Desktop clients for macOS and Windows
- Open-source firmware supporting nRF and ESP32 hardware
- Developer SDKs and Model Context Protocol (MCP) server support
Pricing
Open source. Free and open-source software (MIT); pre-assembled hardware or DIY parts sold separately.
Details
- Type
- GitHub repo
- Category
- AI
- Topic
- AI Wearables
- Open source
- Yes
- Platforms
- macOS, Windows, iOS, Android, Web, Self-hosted
- For
- Developers, quantified-self enthusiasts, and professionals wanting local or open-source continuous capture.
- Alternative to
- Rewind AI, Limitless, Friend (wearable)
- Tags
- wearables, transcription, audio, second brain, open source hardware, flutter
Why it was saved
- @kodjima33: A real open-source alternative to closed AI wearables (Rewind, Avi Schiffmann's Tab), with a build-your-own hardware+software stack.
Related
Turn long-form audio and video recordings into written content assets.
Bill of materials and architecture to build an open-source AI necklace
Speech-to-text API providing word-level timestamps and audio event tagging.
Fast local audio transcription benchmark and Python snippet for Apple Silicon Macs.
Open-source 600M English speech-to-text model with fast inference and timestamps
Wearable AI companion that listens to your day and offers context-aware advice
Updated