DigUp: search your own photos, PDFs, recordings and videos by describing what is in them
DigUp is a free, open-source desktop app that, according to its developer, finds files on your computer by what they show or say. Type "the clause about pets in the lease" or "where they talk about sleep," and it says it opens the PDF page or jumps to that minute of a recording. It runs Google DeepMind's EmbeddingGemma 2, an AI model that maps text, images and sound into one searchable space, entirely on your machine. The developer says nothing you index leaves your computer. For anyone sitting on years of shoots, footage, podcasts or client documents, that could shorten the hunt for one file.
- For
- Detroit creatives and small business owners with large local archives of photos, video, audio, PDFs or documents who want to find specific files or moments without uploading them to a cloud service.
- Cost
- Free and open source (MIT license), with no account, and the developer says it stays free. An optional "buy me a coffee" donation is offered. Requires a one-time 865 MB model download. The developer says it goes online only to download the model and for a daily update check, which can be turned off.
- Time
- Per the developer, the first indexing pass on a Mac takes a few minutes for a few hundred files and can take an hour for a big Downloads folder. On Windows and Linux running on the processor alone, a large photo library can take hours. Setup shows an estimate for each folder, measured on your computer. After that, new files are searchable within seconds.
- Access
- Mac: Apple Silicon with macOS 14 or later. Windows: 64-bit Windows 10 or 11, beta, installs without admin rights, but the installer is not yet signed, so Windows may warn you. Linux: beta, 64-bit Intel/AMD, Ubuntu 22.04, Debian 12, Mint 21 or newer. On Windows and Linux, text inside screenshots is not read yet, and RAW photos, SVGs and old Word .doc files are read only on a Mac. It skips cloud-only iCloud and OneDrive files.
Your moveDownload DigUp from its GitHub releases page, point it at one folder first, and check the time estimate setup shows before indexing a large photo or video library.
Start here