DigUp Uses Google's New Model to Transform Mac Search: Forget the File Name, Still Find a Specific Shot in a Video

Beating AI News Flash: Finding files on a Mac usually requires remembering names or keywords. If you only remember that a zebra appeared in a certain video, or that a podcast episode discussed sleep, traditional file search is of little help.

Independent developer Abdur Rahim's open-source DigUp lets you find files directly by content. Type "the zebra in the video," and it jumps to the shot where the zebra appears; search "the part where they talk about sleep," and it pinpoints the relevant timestamp in the podcast. It can also search photos, PDFs, and even open directly to the page where a specific clause in a contract is located.

DigUp uses EmbeddingGemma 2, released by Google DeepMind on October 6. This model converts text, images, audio, and video into comparable numerical features, allowing a text description to find the corresponding visual or audio content. DigUp pre-indexes selected folders, and during search it can also precisely match content such as invoice numbers and error codes. The entire indexing and search process runs locally on the Mac, with no need to upload private files to the cloud.

DigUp is free and open source, and within less than a week of release it has gained over 3,000 GitHub Stars. Currently it only supports Macs with Apple silicon, requiring macOS 14 or later. First-time use requires downloading a model of about 865 MB and spending time building the index.