Hey everyone, I have been building G1 AI Glasses Companion, an iOS companion app for Even Realities G1 glasses. I know with G2 you have EvenHub which allows the development a lot easier but I only got G1. The goal is simple: make the glasses useful during real conversations without having to pull out the phone.
The app listens from the iPhone/G1 (though not really good), transcribes speech in real time, detects useful questions or moments, and send LLM based answers directly to the G1 HUD.
It is built around live meetings, social interactions, hallway conversations, study sessions, and any situation where quick context in your line of sight is more useful than another screen.
- Live conversation answers with G1 HUD output: start listening, let the app transcribe, and get short answers when a question comes up. answers can be displayed on the glasses with touchpad page scrolling.
- **Customizable HUD Display:**See figure 2.
- RAG based Answer: Upload your documents and it will generate answer based on your docs.
- Project memory: save useful conversation context, project notes, action items, and recurring details.
- Background fact-checking: after an AI answer, the app can keep verification context available without interrupting the conversation.
- Persistent Long Term Memory: Generates long term meory after sessions.
- Provider choice: bring your own key for OpenAI, Anthropic (WIP, not tested), DeepSeek(WIP, not tested), Qwen (WIP, not tested), or OpenRouter-compatible providers.
The app can still be used without glasses for provider setup, transcription testing, history, memories, and HUD previews, but the full experience is obviously best with paired G1 glasses.
Looking for feedbacks and also thinking about provinding this open source:
- Would you use live AI answers during meetings or interviews?
- Is the HUD output more useful as short direct answers, bullet points, or richer context?
- What would you want on the idle dashboard when the glasses are not actively answering?
- Any features you would avoid because they feel too distracting in real conversation?
I can share the TestFlight link with anyone who wants to try it. The project is also open source, so feedback on the implementation is welcome too.