TURN ANY GLASSES INTO AI SMART GLASSES FOR 20 DOLLARS - AND IT REMEMBERS EVERYONE YOU MEET
A 20 dollars ESP32 with a camera and Llama 3 that remembers faces and answers questions about what you saw — clipped onto any pair of glasses.
AIOpen-hardware
- difficulty
- ●●●○○
- time
- a weekend
- license
- MIT
- repo
- repo ACTIVE4,100 stars
●●●○○ · a weekend · MIT · 4,100 stars · repo ACTIVE
WHAT YOU’LL NEED
- 3D printer + filament — printable parts — files are in the repo
- Electronic parts — full list with part numbers in the repo BOM
- Dev board / microcontroller — runs the project firmware
Partner
COMPAREE VERDICT
OpenGlass won a Llama 3 hackathon and the idea is exactly what it sounds like: a XIAO ESP32-S3 Sense (camera plus mic, 15 dollars) in a 3D-printed clip that sits on your glasses, takes a photo every few seconds, transcribes what you say, and lets Llama 3 answer questions like 'what was that person's name' or 'where did I leave my keys'. The whole build is about 20-25 dollars if you already own a 3D printer. The firmware is MIT-licensed and flashes via Arduino IDE. The hard part is not the hardware — it is getting the backend server running on your machine and pairing the mobile app (iOS or Android, both available). The README does not walk you through that; it lists the steps and assumes you know what 'clone the repo and run the server' means. The camera is a tiny fixed-focus module, so image quality is closer to a webcam than a phone — it works for face recognition at conversation distance but will not read small text across a room. The battery (a 3.7V LiPo you solder on yourself) lasts a few hours of continuous use. If you are comfortable flashing ESP32 boards and running a local server, this is a genuinely clever weekend build. If you have never done either, you will spend most of that weekend troubleshooting the software stack, not the glasses.
IN THE REPO
NOT IN THE REPO
- —MIT license — commercial use is fine.
- —3D files for the mount are in the repo, firmware flashes via Arduino IDE.
- —You need a Fireworks AI API key (free tier exists) and a mobile app install to talk to the glasses.
- —The repo includes a backend server you run yourself — no cloud sign-up beyond the API key.
- —The camera is fixed-focus and low-resolution; this is not Google Glass-grade optics.
- —No assembly guide — the README assumes you know how to solder headers and flash an ESP32.
Can I build this?
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
Partner · KickstarterHeyGears G1: 10M+ colors and transparent parts in one print, plus UV printing on flat objects. Figures, parts and labels — no painting.
Videos
OpenGlass Demo - Meta AI LLaMa Hackathon - 12th May 2024
No official build video in the repo or linked from the README. Community walkthroughs may exist but were not verified.
Gallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Read the hardware setup section in the README (Lists the exact XIAO board model and battery specs)
- 2.Flash the firmware via Arduino IDE (Install the ESP32 board package first, then upload the sketch)
- 3.Run the backend server on your machine (Requires Node.js and the Fireworks API key in a .env file)
- 4.Install the mobile app and pair (iOS and Android apps are linked in the repo)
Resources
Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.
- Hardware setup guide github.com
- Firmware instructions github.com
- Backend server setup github.com
- Build video youtube.com
- Project site openglass.ai
Partner · KickstarterCORE POWER BANK S: 0.2" thin, 2.8 oz, 5000 mAh with 22.5W fast charging. Semi-solid battery, carbon fiber, IPX7 waterproof.
KNOWN ISSUES
- The XIAO ESP32-S3 Sense is a specific board variant — the plain ESP32-S3 does not have the camera module. Ordering the wrong one wastes a week and $15.
- The battery needs a JST connector that matches the board's socket — not all 3.7V LiPos come with one, and soldering directly to the pads is easy to get wrong.
- The backend server must be running on the same network as your phone, and the IP address has to be entered manually in the app. If your router assigns a new IP every time, you will re-enter it every session.
- The free Fireworks API tier has rate limits — heavy use (constant photo capture and queries) will hit them in a day or two.
- The camera auto-captures every few seconds by default, which drains the battery fast. You will want to tweak the interval in the firmware for real-world use.
- The 3D-printed clip is designed for standard metal frames. Thick plastic or acetate frames may not fit, and you will need to modify the STL.
Do I need to keep my phone nearby?
Yes. The glasses connect to your phone over Bluetooth, and the phone talks to the backend server. The ESP32 does not have enough memory to run the AI itself.
Can I use a different AI model instead of Llama 3?
Yes — the backend code is modular. You can swap in another model that Fireworks supports, or point it at a local Ollama instance if you want to avoid API costs entirely.
How good is the face recognition?
Good enough for conversation distance in decent light. The camera is low-res, so do not expect it to work across a crowded room or in dim lighting.
Is there a way to turn off the camera but keep the mic?
Not in the default firmware, but the code is open — you could comment out the photo capture loop and reflash.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
The whole thing runs on a $15 board and talks to Llama 3 through your phone — no cloud sign-up beyond a free API key. What would you ask it first?
BasedHardware
BasedHardware built OpenGlass for a Llama 3 hackathon and won. The project is MIT-licensed and actively maintained, with both hardware and software updated regularly.
Keep building — related projects
Same category, picked by how close they are to this build.
NO CAD FILESBuild a talking AI assistant on a breadboard and host its brain yourself
AIOpen-hardware●●●○○moderate
NO CAD FILESA 28.9M parameter language model runs entirely on an eight dollar microcontroller
AIOpen-hardware●●●○○moderate
NO CAD FILESA 28.9M parameter language model runs entirely on an 8 dollars microcontroller
AIOpen-hardware●●●●○advanced
FULL CADThis retro talking head runs a private ChatGPT-style AI fully offline on one Raspberry Pi
AIOpen-hardware●●●●○advanced
DISCLAIMER
- Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
- Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
- Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
CompareeTEAM25d agoedited
Practical notes from our verification: The repo is actively maintained and the MIT license is clearly stated. The hardware list is complete and the XIAO ESP32-S3 Sense is easy to source. The single biggest time sink is not the build — it is getting the backend server and mobile app paired correctly, especially if your router reassigns IPs. The README assumes you know what 'run the server locally' means and does not walk through common networking mistakes. The 3D-printed clip is simple but frame-specific — test-fit before you glue or solder anything permanent. No official build video exists; the repo photos and code are the only guide.