Hey Taby
Taby’s cartoon hand holding a photo

Blog · Smart search ·

Find your photos by what they mean

You remember the moment, not the file name. Ask Taby for “us saying goodbye at the airport” and it can find the photo of you two hugging by the suitcase, even though you never said “hug” and the file is called IMG_2041.jpg. Here’s how that works, right on your own computer.

Coming with Taby 1.0.14

Photo search comes with Taby 1.0.14, the next free update.

Normal search looks for letters

Most search on your computer matches letters. Type “hug” and it looks for files with “hug” in the name or the text. A photo has no words in it, and its name is usually just a camera number. So letter search can’t find it at all.

Search by meaning looks for ideas

Search by meaning, also called semantic search, works differently. A small AI model looks at each photo and gives it a spot on a huge map of meaning. Photos about similar things land close together: beaches near beaches, birthday cakes near birthday cakes.

When you search, your words get a spot on the same map. “Saying goodbye at the airport” lands right next to a photo of two people hugging at the gate. Taby shows you the photos closest to your words. No tags, no folders, no exact words.

The short version

Search by meaning turns photos and words into points on one map. Points close together mean similar things. Taby finds the photos closest to what you asked for.

One model for words and pictures

The trick is that words and pictures have to share the same map. Taby uses EmbeddingGemma 2, a small open model from Google DeepMind that was made for exactly that, so a sentence can find a photo.

The same model now runs the rest of Taby’s search too, for your notes, tasks and reminders, and it made that better. In Taby’s own test, it found the right note first 101 times out of 102, up from 97 with the model before.

It all happens on your computer

  • You choose the folders. Finding files stays off until you say yes.
  • Taby reads in the background, one photo at a time, newest first. It waits while your laptop runs on battery, while Taby is answering, or when your computer is busy.
  • Your photos aren’t uploaded to be searched. The map lives on your computer, and the little previews Taby shows are made by your computer and never uploaded.
  • One small download. The search model is a one‑time download of about 540 MB.
  • Free, on either brain. It works with the free Local Brain and with Taby Cloud alike.

In your language, and honest about misses

You can ask in any of Taby’s 14 app languages. We tested it on a public set of 1,000 photos, asking for 20 of them by a short description in each language. The right photo came first 10 to 13 times out of 20, was in the top five 15 to 19 times, and was always in the top 20.

So it’s good, not magic. If the first photo isn’t the one, the right one is usually close by. A little more detail helps: who’s in it, where it was, what was happening.

How to try it

  • “Find the photo of us saying goodbye at the airport”

    Taby shows the closest photos as small pictures. Open one, or show it in Finder or File Explorer.

  • “Find the picture with the white dog”

    Say what’s in it. The file name doesn’t matter.

  • “That photo of my birthday cake”

    Talk or type, in your own words. Taby searches the folders you picked.

  • “Show me the sunset photos from our trip”

    Ask the way you’d ask a friend. Taby looks for what fits.

Prefer typing? Press ⌘K on a Mac or Ctrl+K on Windows and type what you remember. The search box finds pictures by what they show, next to your notes and files.

Photos are one part of smart search, which also finds your notes, tasks, reminders and files by just asking. Want to know what else runs on your own computer? Read what local AI can do.

Find it by just asking

Download Taby free for Mac or Windows. Photo search comes with the 1.0.14 update, and Taby updates itself.