Windrecorder: A Local-First Screen Memory Tool for Windows

You know that feeling when you saw something on your screen three weeks ago—a chat message, a settings panel, a price—and you can't find it again no matter how hard you search? Scrolling through browser history and file explorers only gets you so far. What if you could just rewind your screen like a DVR and search through everything you've ever looked at? That's exactly what Windrecorder aims to give you, and it does it entirely on your own machine.

Windrecorder is an open-source Windows app that records your screen continuously, indexes what it sees, and lets you search through that footage using OCR text or image descriptions. Think of it as a self-hosted alternative to tools like Rewind or Copilot Recall, but with one key difference: nothing leaves your computer. No cloud upload, no internet connection required, no data going anywhere except your own hard drive.

What It Does

At its core, Windrecorder is a personal memory search engine. It captures your screen in the background—either multiple monitors, a single display, or just the active window—and compresses that footage to keep file sizes manageable. It's designed to run continuously without hogging your system resources, so you can forget it's there until you need it.

But recording alone isn't useful. The real magic is in the indexing. Windrecorder doesn't just store video; it detects when your screen changes and extracts the meaningful bits: OCR text, page titles, and browser URLs. All of that metadata goes into a database, which you can then query through a full web interface. Want to see what you were looking at when you read about "Rust lifetimes" last Tuesday? Search for that phrase and Windrecorder will pull up the relevant screen moments.

The project also includes some thoughtful automation. You can set custom skip conditions—like ignoring specific window titles, process names, or periods of screen stillness—so it's not wasting storage on static screens. It'll even maintain itself, cleaning and compressing video when your computer is idle. And if you're into data about your own behavior, it generates activity statistics, word clouds, timelines, and scatter plots from your usage patterns.

Under the hood, it's built with Python and runs on Windows. The OCR engine is pluggable: you can stick with Windows' built-in recognition or swap in third-party engines like RapidOCR, WeChat OCR (which has notably high accuracy for Chinese and English), or Tesseract OCR for over 100 languages. The interface itself supports Simplified Chinese, English, and Japanese, and the project welcomes translations.

Why It's Cool

The most compelling thing about Windrecorder is the privacy stance. In an era where every "helpful" feature seems to phone home, this tool runs completely offline. You own all your data, period. That's not a marketing line—it's a design constraint that shapes the whole project.

• It's a Rewind alternative that actually runs on Windows. Rewind is Mac-only, and Copilot Recall has had its share of controversy. Windrecorder fills that gap for Windows users who want similar functionality without the platform lock-in.

• The indexing strategy is smart. Instead of running OCR on every single frame (which would be wasteful), it only processes changed scenes. That's a practical engineering decision that keeps resource usage low and the database clean.

• The AI summarization is opt-in and local. It supports tag summarization using LLMs, but since everything runs locally, you're not sending your screen data to some API. That's a nice middle ground for people who want AI assistance without the privacy trade-off.

• The OCR engine flexibility is genuinely useful. Different OCR tools excel at different languages and scripts. Being able to swap engines based on your needs—or contribute your own—is the kind of extensibility that makes open source shine.

• It's honest about its maturity. The README carries a warning that this is early-stage software. That's refreshing, and it sets the right expectations for anyone diving in.

How to Try It

Getting started requires a few pieces of software, but nothing exotic. Here's the gist from the README:

1. Install ffmpeg. Download ffmpeg-master-latest-win64-gpl-shared.zip from the FFmpeg builds page and extract the bin files to C:\Windows\System32 (or any directory in your PATH).

2. Install Git. The standard Windows installer works fine—just keep clicking next.

3. Install Python 3.11. This is important: Python 3.12 is not supported yet, so make sure you grab 3.11 specifically. During installation, check the box for Add python.exe to PATH.

4. Clone and run. Navigate to your chosen install directory and clone the repository from GitHub, then follow the setup instructions in the README to launch the app.

The project is under active development, so if you hit snags, the maintainers encourage filing issues or starting discussions. You can find the repository and full installation guide at github.com/yuka-friends/Windrecorder.

Final Thoughts

Windrecorder is a promising tool for anyone who spends a lot of time in front of a Windows machine and wishes they had a photographic memory of their own screen. It's not polished consumer software yet—the early-stage warning is real—but the core concept is solid and the local-first approach is the right call. If you're a developer who tinkers, this is also a great project to contribute to, whether that's improving OCR support, refining the UI, or adding translations. For now, it's a genuinely useful utility with a clear roadmap and a sensible privacy posture. Worth keeping an eye on.

Follow @githubprojects for more developer tools and open source projects.