After unveiling Recall AI at the Copilot+ PC event, Microsoft received criticism from the Windows community regarding privacy and security issues. Microsoft responded by making the Recall feature opt-in by default and introducing “just-in-time” decryption for enhanced security. The rollout of the Recall feature has been postponed by Microsoft, with plans to offer a preview within the Windows Insider Program before the official release.
This implies that Recall AI will not be accessible on Copilot+ PCs set to launch on June 18. However, for those interested in exploring an open-source alternative to Windows Recall, OpenRecall is worth considering. Compatible with various PC types (x86, Windows, Linux, macOS), OpenRecall does not necessitate a separate NPU.
OpenRecall Functions on Intel and AMD PCs Without an NPU
OpenRecall, an open-source project on GitHub, captures screen screenshots and conducts analysis through small AI models on the device. This mirrors Microsoft’s Recall setup but is compatible with x86 Intel/AMD PCs and does not rely on an NPU.
OpenRecall assesses texts and images, saving the embeddings in an SQLite database on your PC. It boasts a user-friendly web UI for semantic searches and activity retrieval.
One drawback is the storage location for screenshots and the DB file at, C:\Users\username\AppData\Roaming\openrecall accessible to anyone without encryption. I managed to access the DB file using SQLite Browser, highlighting OpenRecall’s current lack of robust privacy and security measures.
The project indicates future plans for encryption implementation, along with upcoming features like app filtering, audio transcription, and enhanced visual search capabilities.
In general, I appreciate the capability for users to utilize a Recall-like feature on non-Copilot+ PCs without the need for new hardware. Nevertheless, the absence of encryption remains a significant issue. If you wish to try OpenRecall on your PC, kindly refer to the guide provided below.
How to Run OpenRecall on Any PC
- Ensure Python, Pip, and Git are installed on your PC.
- Subsequently, launch the Terminal and execute the single-line command provided.
python -m pip install --upgrade --no-cache-dir git+https://github.com/openrecall/openrecall.git
- Then, proceed to execute the following command to initiate the OpenRecall application.
python -m openrecall.app
- Access the localhost address in a web browser to view all your activities displayed with a slider, reminiscent of Windows recall. Each activity includes its respective date and time.
- Subsequently, I conducted a search for the “India vs USA” match on Google and later in OpenRecall, where a simple “cricket” query generated the corresponding Google search results, showcasing semantic search functionality.
Please note that OpenRecall currently lacks the capability to extract texts using OCR and does not excel in visual searches. It may require a compact vision model for on-device object detection, potentially straining the PC’s resources. During my testing, OpenRecall utilized approximately 500MB of RAM, with Python’s CPU usage hovering around 20%.
This outlines how you can leverage an open-source alternative to Windows Recall on your existing Windows PC. If you encounter any challenges, feel free to reach out in the comments section below.
