Search

Alt+Space

Press Alt Space from anywhere, even from inside a program, and type. Search is instant: results come in as you type, with a preview of the selected one (size, date, what a folder contains). Esc closes.

If another program already uses Alt Space (PowerToys Run does, for example), Warpstead uses Ctrl Alt Space instead.

Results are split by type. All puts the top result on the left, with the folders under it, and on the right the images as thumbnails, then documents, video and audio, and everything else. Each section shows the first few; Show all opens its tab. The tabs at the top (Folders, Images, Documents, Video and audio, Other) show one type at a time, with the count next to the name. In Images the thumbnails are grouped by folder.

Search looks at the names of files and folders, not at what’s written inside them; photos can also be found by what they show, if you turn it on (see below). It works offline, without administrator rights, and nothing leaves your PC.

Which folders

By default Warpstead indexes Desktop, Documents, Downloads, Pictures, Videos and Music. Change the list in Settings → Search: Add folder… to index another one, Add exclusion… to skip a folder inside them. The change applies at once.

Folders that sit directly in your user folder, next to Documents and Desktop (a project in C:\Users\yourname\Project, for example), are found by name even when they aren’t on the list: they show up at the top marked Not in search. Select one and Add to search puts it on the list; from then on the files inside are found too. Hidden folders like AppData, and those starting with a dot, stay out.

Network folders can be indexed if they’re mapped to a drive letter. While indexing runs, a notice in the corner says so, and Settings shows the state of each folder.

Photos by content

You can find a photo by describing what’s in it: “birthday cake with candles”, “dog on the beach”. Turn it on in Settings → Search with Search photos by content, or with Turn on content search in the search window. The first time, Warpstead downloads the model that reads the photos, once; its size is written under the switch.

The model then goes through the photos in your indexed folders one by one and saves a fingerprint of what each one contains. Search compares your words with the fingerprints, so every photo is read once and after that only new ones are. On a big archive the first pass takes hours: with a graphics card it reads roughly 6 to 15 photos a second, without one it’s slower and Settings tells you so. The notice in the bottom right shows how far along it is. On battery, or when the PC is busy, it pauses and picks up again later. All of this happens on your PC and the photos aren’t sent anywhere.

In the Alt Space window, in All and in Images, the photos found for what they show come after the results by name, under a By content row with the reading status, as a grid of thumbnails, closest match first. A long name fits on two lines; select the thumbnail to read it in full. The folder icon on a thumbnail opens the folder that holds the photo. A photo that matches both by name and by content shows up once, among the names, and the counter at the top adds the two parts: 3 + 12 by content. The arrows, Enter, Ctrl Enter and Ctrl T work as on the other results.

It only covers photos. Photos that can’t be decoded are skipped and won’t turn up by content. OneDrive files that are online-only aren’t read, so they don’t all get downloaded, and a disconnected network folder waits until it’s back.

Delete photo fingerprints… in Settings frees the space they take. The model stays on disk; if you turn content search back on, every photo is read again.

Filters

Type key:value in the box to narrow the results. You can mix filters and words: ext:pdf invoice.

kind: also takes the Italian words. An unknown key is searched as plain text. With the box empty, a few filters are suggested: click one to use it.

Tags

Tags are words you attach to files: by hand with Ctrl T in the results, or written by the AI (Tag with AI in a Folder node). Find them with tag:.