Why filename search fails
Names like IMG_7843.png tell you nothing months later. Browsing folders to find one frame wastes real time, and tags only help if you set them up first.
You remember the photo. You cannot remember the filename, the folder, or the day you took it. PhotoCore lets you type what you remember, like 'red umbrella on a rainy street', and brings up the matching images in milliseconds. There are no tags to set up first, and nothing is uploaded.
PhotoCore reads the content of every image and turns it into a compact vector. When you type a phrase, it matches your words against that meaning and ranks the closest images first. A short query like 'business model diagram' or 'sunset at the sea' is enough.
Names like IMG_7843.png tell you nothing months later. Browsing folders to find one frame wastes real time, and tags only help if you set them up first.
Describe the scene, the object, the mood, or the text on a cover. PhotoCore finds the image even when its name and folder say nothing about it.
The local SigLIP 2 model understands queries in more than 100 languages, so you can search the way you think, with no setup.
Every vector is computed and stored locally. No pixel is uploaded, and search keeps working with no connection.
Searching this way is often called semantic search. For the short version of how it works, read how semantic photo search works. The same engine powers Constellation, which turns a search into a visual map of groups.
Download PhotoCore free, index up to 200 images, and search them by description. Full access is a one-time purchase.