(ProjectMnemonic)
This project is a lightweight, portable app designed to record everything happening on your screen. It's like having your own personal screen recorder 😬 Not perfect yet, but cool if you need to capture and analyze your screen without any strings attached. While it’s not quite Microsoft Recall, it still offers a similar experience (this is the first release).
Why? Microsoft built something the world didn’t really ask for. That was my first thought, and to be honest, I’m still torn on Recall. So, I decided to recreate it myself—giving everyone the chance to explore the concept without needing a Copilot+ PC. Consider this a sneak peek at what Microsoft is pushing with Recall on Copilot+ PCs.
And guess what? It’s already on your 24H2 system, deeply embedded into Windows and Explorer (though disabled by default). Recall is far from perfect, and this article paints a pretty accurate picture of its ups and downs: Computerworld Hands-on with Windows Recall. I won’t dive into all the flaws and concerns here—that’s a discussion for another time. My goal? Just to show you what Microsoft is aiming for with Recall. I’ll have more thoughts on this later.
Technical Details: The app uses Hugging Face’s BLIP model for cloud processing but also works offline, with all OCR tasks handled locally by Tesseract.
Currently, I’m running this project under the name Winrecall(4All).
If you don’t want to invest in a Copilot+ PC (the only way to access Microsoft Recall, since those devices have the necessary NPUs for AI-intensive tasks), this open-source alternative lets you test the waters—without the hefty price tag.
- Capture & Process: The app captures your main screen at regular intervals.
- AI Descriptions: Each screenshot is analyzed, and a detailed text description is generated.
- Search & Retrieve: Screenshots and their descriptions are indexed in a local SQLite database, making them searchable by keywords.
- Controls: Hit "Capture" to start, "Stop" to pause, and manage your saved images through the database settings.
Run WinrecallApp.exe, then click the "Setup Environment" button in the bottom left corner.
- The app will automatically install Python (if needed) and set up all the required dependencies.
- Once everything’s set up, restart the pc and you’re good to go!
If you prefer to install everything manually, here are the steps:
- Install Python (make sure to check "Add Python to PATH" during installation).
- Open a terminal and run:
python -m pip install --upgrade pip
pip install transformers torch pillow
(Optional) Run BLIP Image Captioning Model Locally
Fork and clone the BLIP model from Hugging Face:
git clone https://huggingface.co/Salesforce/blip-image-captioning-base
In Winrecall/describe_image.py file, change lines 9-10 to load the model locally:
processor = BlipProcessor.from_pretrained("./blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("./blip-image-captioning-base")
This will use the locally downloaded files instead of fetching the model from the cloud.