TimeScroll: An Open-Source Local Alternative to Rewind for Mac
A privacy-focused screen recording tool optimizing CPU and storage usage via accessibility APIs.
1Background Screen Recording
TimeScroll is a new open-source macOS application designed to record screen activity in the background. It allows users to scroll through their daily history or search for specific visual content. The project emerged following the shutdown of Rewind, aiming to provide a more efficient and polished alternative for local data tracking.
2Direct Mode Innovation
Unlike standard apps that rely on heavy OCR processing, TimeScroll utilizes the macOS Accessibility API to extract text. This method, termed Direct mode, captures exact text strings and positions without misread characters. It significantly reduces CPU load, as it avoids the computational intensity typically associated with constant optical character recognition on screenshots.
3Handling Complex Applications
Implementing Direct mode required overcoming challenges with Chrome and Electron applications, which do not always expose an accessibility tree by default. The developer manually sets the AXManualAccessibility attribute to force tree creation. Additionally, custom logic traverses these complex hierarchies within strict time limits to ensure efficient and accurate text data capture.
4Performance and Efficiency
The application demonstrates high performance, requiring only 9ms of CPU time for initial frame reads, dropping to 0.4ms for subsequent ones. On an M4 chip, the app maintains an average of 1.1% CPU usage and 150-250MB of memory. This represents a substantial improvement over legacy tools that previously consumed significantly higher system resources.
5Storage Optimization Techniques
TimeScroll uses HEVC video encoding to manage storage, creating one segment file per minute. By implementing constant quality settings and specific keyframe intervals, the developer reduced storage requirements to approximately 2GB per month. Advanced deduplication and compression techniques further minimize the footprint of redundant text data across multiple captured snapshots.
6Local AI Summarization
The app features a summarization tool that processes user activity over selected timeline ranges. By dividing usage into sessions using heuristics and selecting representative screens via embeddings, it generates written summaries through a local Ollama model. This ensures that sensitive user data remains private and is processed entirely on the local machine.
7Technical Implementation Challenges
The developer spent significant effort tuning the encoding process to balance quality and file size. On text-heavy frames, the chosen approach reduced frame sizes from 84 KB to 20 KB. These optimizations were necessary to prevent the storage bloat commonly seen in screen-recording software, which often leads to rapid disk space exhaustion.
8Future Development Outlook
TimeScroll represents a shift toward lightweight, privacy-conscious productivity tools on macOS. By prioritizing local processing and efficient API utilization, it addresses the primary pain points of battery drain and storage overhead. Future iterations will likely focus on refining the Direct mode and expanding the capabilities of the local AI-driven summarization features.