What counts as open source smart glasses?

Open source can refer to hardware schematics, firmware, phone companion app, and cloud components. Omi (BasedHardware) publishes pendant + Glass hardware, firmware, app and cloud under MIT. MentraOS publishes its Smart Glasses OS/runtime (MIT) for Even Realities G1, Vuzix Z100 etc. CyanBridge publishes its Android companion source on GitHub and keeps data paths inspectable.

Omi vs CyanBridge vs MentraOS — which should I choose?

Omi is a pendant-first wearable for always-on memory (screen + conversations, MIT, Ollama via localhost:11434). MentraOS is an OS/runtime to build one app that runs on multiple display glasses. CyanBridge is a vendor-independent bridge that lets you use glasses you already own (HeyCyan, EyeVue, MYVU, experimental Ray-Ban Meta) toward private, local-first AI with self-hosted options.

Can I run AI offline without cloud?

Yes, where hardware allows. Omi supports offline/transcription modes, MentraOS can run on-device apps, and CyanBridge supports local LLMs (Qwen, Gemma), self-hosted Ollama at http://localhost:11434, Groq/OpenAI-compatible endpoints, and explicit local-first routing so cloud is optional and quoted before use.

Open source · MIT · Local-first

Open Source Smart Glasses — Build Private AI on Glasses You Already Own

Open source smart glasses are about control, not only price. This page compares the leading MIT stacks — Omi, MentraOS, OpenGlass — and where CyanBridge fits as a private, vendor-independent alternative to cloud-only AI glasses like Ray-Ban Meta.

Project	What it is	License & repo	AI routing	Best for
Omi	Wearable pendant, Omi Glass, desktop/mobile/web memory. 300k+ professionals. Always-listening transcription.	MIT — github.com/BasedHardware/omi. Docs: docs.omi.me.	Groq / OpenAI / Ollama localhost:11434.	Second-brain memory across work + personal.
MentraOS	Smart Glasses OS + Runtime. One app runs on G1, Z100 etc. Display, mic, camera, speakers unified API.	MIT — github.com/Mentra-Community/MentraOS (2,326 stars).	Phone-side runtime, multiple apps simultaneously, private & inspectable.	Builders who need cross-glasses compatibility.
OpenGlass	DIY <$25 hackable glasses (XIAO + camera). Hack any frames.	MIT — github.com/BasedHardware/OpenGlass (Terryxxx fork, ~2025).	Same Ollama pattern as OmiGlass.	Makers & students.
CyanBridge	Vendor-independent bridge for glasses you already own toward private, local-first AI.	Android companion public on GitHub (Alternative-HeyCyan-App-and-SDK). Model Studio separate.	Local LLM (Qwen/Gemma) → self-hosted (Ollama/Groq) → optional hosted (quoted).	Privacy, accessibility, personal datasets.

Public source you can inspect today

CyanBridge Android companion: github.com/FerSaiyan/Alternative-HeyCyan-App-and-SDK (public). Model Studio remains locally maintained and is not claimed as publicly available — the site now distinguishes Public Android source from Source and Licensing (see /privacy and /terms). For OEM work see Omi Glass announcement (XIAO + camera, longer battery than Meta Ray-Bans, announcement video via x.com/kodjima33).

What open source means for smart glasses

Hardware open (schematics, printable frames) — e.g., OpenGlass <25$ DIY with XIAO + camera, IAm5tillLearning OpenSource-Ai-Glasses (RV1106B, Linux, MIT).
Firmware/app open (phone companion, RTSP streaming, BLE) — e.g., Omi (nRF5340 + nRF7002 Wi-Fi 6 dual mics), Omi Glass battery-first camera + XIAO.
OS open (cross-glasses runtime) — e.g., MentraOS (MIT, 2,300+ stars) for Even Realities G1, Vuzix Z100, shared Display/Mic/Camera API.
CyanBridge angle: public Android companion source (Alternative-HeyCyan-App-and-SDK) plus local-first data paths; Model Studio is maintained separately and not claimed as publicly available.

Self-hosted and offline AI: Ollama, Groq, OpenAI where you decide

OmiGlass default: Groq key, OpenAI key, or Ollama URL http://localhost:11434/api/chat for local hosting — pattern CyanBridge also supports for self-hosted inference.
Local vs cloud is a routing choice, not a belief statement. CyanBridge routes to phone/desktop/self-hosted when task fits, and only to an explicitly selected hosted provider when local hardware is insufficient.
Also called private AI glasses, offline AI glasses, self-hosted AI glasses — same goal: data stays with the user unless the user approves a quoted cloud job.

Privacy model: always-listening pendant vs display glasses bridge

Omi markets always-on wearable memory (conversation, screen, summaries, action items). That maximizes recall but requires deliberate permission and retention design. Display glasses (Even Realities G1, Z100, CyanBridge-bridged devices) maximize glanceable output. CyanBridge bridges the latter category: keep capture minimal, permissions explicit, and storage personal where possible.

Where CyanBridge fits in the open ecosystem

If you already own HeyCyan, EyeVue, Meizu MYVU, Ray-Ban Meta (experimental) or other bridged frames, CyanBridge lets you inspect exports, keep datasets local, and run Qwen/Gemma or other models on hardware you control. It is complementary to Omi (pendant) and MentraOS (OS) — choose the form factor and stack that matches your workflow, or combine them.

Ray-Ban Meta alternativePrivacy-focused, private comparison vs cloud-only AI glasses.Local AI for smart glassesFour-layer architecture: glasses → phone → local/self-hosted → optional cloud.Compatibility matrixCheck Supported / Experimental / Research status before planning a workflow.Qwen vs Gemma for local AIModel selection framework for phone, desktop or local server hardening.

Start from the glasses you already own

Check device compatibility, then explore local-first and open-source bridge paths.

Continue