Show HN: LarpLess, open-source study tracker detecting distractions locally
LarpLess is a macOS/Windows app that checks your screen and camera with local AI to determine what you are studying or if you are distracted. Gives you a focus percentage at the end, with a breakdown of what you did during your session (Math, Physics, etc.)
Privacy is a big concern for an app like this, so it is open-source (AGPL 3.0) and uses locally run open-source models. It doesn't connect to the internet unless you want to check for updates or enable the friends feature (Disabled by default. You can make an account, add friends and see their shared study sessions/share your study sessions).
For the camera, it uses an object detection model, YOLO11n, to detect if there is a person and/or a phone at 1fps. For the screen, there are 2 options: the one we did first is reading window titles + a VLM on Ollama (default is qwen3-vl:2b), simple but power/RAM hungry. The one my friend added later can run on an average laptop easily. It takes less than a second for a verdict. It scores window titles + OCR text on the screen with MiniLM text embeddings and scores the screenshot with image embeddings using SigLIP 2. SigLIP 2 decides if the user is distracted, and MiniLM decides what they are working on. You can run the classifier-based mode on any modern laptop. If you want to use the VLM mode, you need a good GPU with enough VRAM (3GB+) to hold the model in it.
You can try it out by downloading the macOS version from GitHub (https://github.com/AlperErg/larpless/releases) or the Windows version from the Microsoft Store (https://apps.microsoft.com/detail/9nrvvphtmr9x).
I found it helpful for studying. I hope that you will as well. I'd appreciate it if you can give it a go and kindly share your thoughts.
p.s. My friend, who also works on this with me, posted about it before but without a description and not on Show HN, so I decided to post on Show HN with a good description.
There are no comments yet.