My labmate and I built the image comparison tool we needed for our group. It runs offline, but can generate shareable links for our PI and collaborators.

Researchers have publicly released TwinLens, a free image comparison tool built to solve a specific pain point in AI and computational research labs. The tool was created by lab members who needed a better way to visually compare outputs from different training steps, check for quantization errors, and benchmark results against other work. Frustrated with standard image viewers not designed for side-by-side analysis, they developed a local-first application that prioritizes speed, privacy, and collaboration. The core innovation is its hybrid approach: it runs entirely offline on the desktop or as a web app, but can optionally generate shareable links that preserve the exact view state for collaborators.

Technically, TwinLens is designed for the research workflow. When a shareable link is generated, images are temporarily uploaded to a server and auto-deleted after two weeks, while the link captures the precise zoom level and comparison settings. This eliminates the need to send large ZIP files and ensures collaborators see exactly what the researcher intended. A key feature for publication is the one-click 'Snapshot Export,' which creates clean, labeled comparison images ready for slides or paper submissions. The tool is now available for free on Mac, Windows, and Web, with the developers promising quick updates as they use it daily in their own lab. Its release highlights a growing trend of researchers building and sharing the specialized tools they need, rather than waiting for commercial solutions.