DigUp is a free, open-source application that helps Mac users find files by describing their contents instead of remembering filenames. The tool searches across photos, documents, PDFs, audio recordings and videos, making it easier to retrieve information stored on a computer. It uses Google’s EmbeddingGemma 2 artificial intelligence model and processes searches locally, without uploading personal files to cloud services.
The application addresses a common problem for computer users who accumulate large collections of digital files. Finding a particular screenshot, document, or audio recording can be difficult when you don’t know the filename or what it contains. DigUp simplifies this process by letting users search in plain language, including descriptions of images, topics discussed in recordings, and information in documents.
DigUp uses natural language to locate files
DigUp is available for Apple Silicon Macs running macOS 14 or later. Users can select the folders they want the application to index before searching their contents through natural language queries. Rather than entering a filename or navigating through several directories, they can describe what they remember about a file.
What if Spotlight actually understood the meaning inside your files? DigUp is an open-source macOS menu bar app powered by EmbeddingGemma 2 that brings 100% local, multimodal semantic search to your Mac. Instead of searching for file names, you can use natural language to: 📸 Find specific scenes in your photos ⏳ Jump to the exact timestamp in audio or video 💻 Search through your codebase by intent Everything runs locally in a single vector space.
— Google Gemma (@googlegemma) October 13, 2026
For example, a search for a dog near water could identify relevant photographs in a collection. Someone looking for a particular discussion in a podcast could enter a phrase such as “where they talk about sleep”. The application can then locate the relevant section of the recording, helping users find information without listening to an entire episode.
The same approach applies to documents and screenshots. A user searching for a specific clause in a lease can find the relevant PDF page. At the same time, a query like “payment declined error” can identify a screenshot containing the corresponding message. These capabilities could help people who regularly work with contracts, research materials, meeting recordings, and large media libraries.
DigUp also supports code searches through an optional feature. This lets developers find functions by purpose rather than exact names. For example, a developer who remembers what a particular function does but cannot recall its identifier could use a description to locate the relevant code.
These features run on Google’s EmbeddingGemma 2, a multimodal embedding model with 740 million parameters. Instead of treating text, images, audio and video as entirely separate search categories, the model represents their content in a shared space. This allows the application to identify relationships between a user’s description and the meaning of the indexed material.
DigUp combines this semantic approach with exact-word matching. This distinction matters because descriptive searches aren’t always enough to locate precise information. An invoice number, error code or specific phrase may be easier to find through an exact match, while a general description benefits from semantic search. Combining both methods gives users more flexibility when searching different types of files.
Local processing keeps personal files on the Mac
Privacy is one of DigUp’s main features. According to the developer, the application processes searches locally and keeps indexed files on the Mac rather than sending them to an external cloud service. Users don’t need to create an account, and the application doesn’t require telemetry.
The EmbeddingGemma 2 model is about 865MB and downloads once. After installation, the model can run on the computer, allowing users to search indexed content without repeatedly uploading files to an online service. This may appeal to people who handle sensitive documents, personal photographs, private recordings or confidential work materials.
Local processing also gives users greater control over which content becomes searchable. DigUp requires them to select the folders they want to index, rather than automatically searching every file on the computer. This approach lets users focus on specific directories, but it also means files outside the selected folders won’t appear in searches.
However, the application has some limitations. Initial indexing can take considerable time, particularly for users with extensive photo, video, or audio collections. The application must process the selected material before it can return relevant results, so the time required may depend on the size and type of the files being indexed.
Hardware compatibility is another consideration. DigUp currently requires an Apple Silicon Mac running macOS 14 or later, which excludes older Intel-based Mac computers. Users also need enough storage space for the downloaded model and the additional data generated during indexing.
Despite these requirements, running the search model locally offers a different approach from cloud-based AI services. Users can benefit from meaning-based file discovery without sending their personal collections to an external platform. The trade-off is that the initial setup and indexing process may require patience, especially on computers containing large amounts of media.
DigUp offers an alternative to traditional Mac search tools
Apple’s Spotlight already helps Mac users find applications, documents and other files through system-wide search. Other semantic-search tools and local knowledge-retrieval applications also aim to make stored information easier to access without relying entirely on filenames. DigUp enters this established category with a particular emphasis on searching across different media types through a single interface.
Its ability to link descriptions to specific content could help people whose files contain information that is hard to represent in a filename. Photographers, researchers, developers, and professionals who regularly record meetings or interviews may benefit from searching by subject matter rather than folder structure.
The application’s support for audio and video adds another dimension to file discovery. Instead of simply identifying a recording, DigUp can help users locate a relevant moment within it. Similarly, its ability to identify relevant PDF pages could reduce the time spent scrolling through lengthy documents. These features make the application focus on finding information within files rather than merely identifying where those files are stored.
The combination of semantic and exact-word searches is another distinction. Natural language helps when users remember the meaning but not the exact wording. Exact matching remains valuable when a search depends on a specific number, code, or phrase. Supporting both approaches could make DigUp useful across a wider range of everyday searches.
Nevertheless, DigUp’s practical value will depend on how well it handles different collections and search descriptions. Indexing time, hardware requirements and the need to select folders manually may influence whether it suits users with extensive digital libraries. Its current Apple Silicon requirement also limits its potential audience compared with tools available across a broader range of computers.
DigUp is available as a free, open-source project on GitHub, giving users access to its development and source code. For Mac owners who frequently forget filenames but remember what their files contain, the application offers an alternative way to organise and retrieve digital information. Its focus on local processing may also appeal to users who want AI-assisted search without relying on cloud-based file analysis.
