GitHub radar
An AI that reads photos and video right on your laptop
LFM3-VL-2B is a model that looks at your photos and video and answers questions about them, running locally with no server or subscription. Install it once and ask it about screenshots, charts, or video clips offline.
A small AI model that looks at photos and video and answers questions about them — right on your laptop, no internet needed.
When it helps
Useful when you need to quickly read a screenshot, a device photo, a chart, or a short video without a subscription or internet. Not built for coding or long conversations — the creators say so themselves.
Pros
- Runs on an ordinary laptop with 8GB of memory, no server needed
- Understands video, not just photos — rare for its size
- Reasons step by step, so it handles charts and counting objects
- Beats every similarly sized model by at least 8 points in tests
Cons
- Not built for coding or long conversations — the creators say so
- The compressed file is about 1.2GB — needs space and download time
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) in your project folder
- 2Send the agent the text from the block below
- 3Wait while the agent downloads the model and runs it through llama.cpp or LM Studio
- 4Ask the model a question about a photo or video in the chat that opens
- 5Check the answer and make sure you have 8GB of memory and 1.2GB of free space
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the LFM3-VL-2B model from Liquid AI locally for me: https://huggingface.co/LiquidAI/LFM3-VL-2B — grab the ready-made compressed (GGUF) build, run it through llama.cpp or LM Studio, and show me how to ask it a question about a photo or video.
A regular laptop with 8 GB of RAM or more — the compressed version of the model takes about 1.2 GB.
Open on Hugging Face▌ More finds