Upon unveiling the Recall AI at the Copilot+ PC gathering, Microsoft found itself under fire from the Windows community due to concerns surrounding privacy and security. After much deliberation, Microsoft has now made the Recall feature an opt-in by default. Moreover, they have introduced “just-in-time” decryption to enhance security measures. However, Microsoft has taken a step back and decided to postpone the Recall feature, opting to preview it first in the Windows Insider Program before rolling it out officially.
This development implies that Recall AI will not be accessible on Copilot+ PCs shipping on June 18. Nonetheless, if one seeks to explore an open-source alternative to Windows Recall, OpenRecall is worth a glance. Compatible with various PCs (x86, Windows, Linux, macOS), OpenRecall operates sans the requirement of a separate NPU.
OpenRecall, an open-source endeavor available on GitHub, captures screen shots and runs analyses through modest AI models right on the device. This methodology mirrors Microsoft’s Recall implementation. Notably, OpenRecall can be employed on x86 Intel/AMD PCs, eliminating the need for an NPU.
Venturing further, OpenRecall conducts analyses on texts and images, storing the embeddings in an SQLite database located on the user’s PC. With a user-friendly web UI, users can undertake semantic searches and revisit past activities.
While OpenRecall appears promising in its functionality, a glaring catch emerges: the storage of screenshots and the DB file in C:\Users\username\AppData\Roaming\openrecall is susceptible to unauthorized access. Lacking encryption at present, the DB file could be opened using SQLite Browser, underscoring OpenRecall’s inadequacies in terms of privacy and security.
To assuage such concerns, the project promises encrypted features in forthcoming releases. Additionally, plans include app filtering, audio transcription, and enhanced visual search capabilities in the future.
OpenRecall offers a viable solution for users to experience a Recall-like feature on non-Copilot+ PCs, sparing the need for new hardware. Nevertheless, the absence of encryption remains a key cause for apprehension. For those keen on testing OpenRecall, a concise guide follows below.
In order to run OpenRecall on any PC, ensuring Python, Pip, and Git are installed is a prerequisite. Following this, executing the prescribed command in the Terminal allows for the installation process to unfold seamlessly.
Once successfully installed, initiating the OpenRecall app is a straightforward affair, requiring the entry of a single command. Accessible via a web browser, the localhost address showcases user activities complete with a slider akin to Windows Recall, complete with time and date stamps for reference.
A semantic search feature unfolds as users delve deeper into OpenRecall’s capabilities. Albeit OCR functionality isn’t incorporated, and visual searches exhibit room for improvement, OpenRecall’s operation maintains a relatively light footprint on resources. Consuming around 500MB of RAM and 20% of Python’s CPU usage, OpenRecall presents a viable alternative to Windows Recall for existing PC users.
Embrace OpenRecall as a beacon of innovation, a testament to the open-source community’s ingenuity. Should any concerns or queries arise, do not hesitate to voice them in the welcoming embrace of the comment section below.
Support our work ❤️
If you enjoyed this article, consider leaving a tip to help us keep publishing great content.
