Skip to content

Quick Start

Get from nothing to a labeled frame in about five minutes.

There is a guided tutorial built into the app

Click Start Tutorial in the menu bar and the app walks you through the workflow, highlighting the exact control to click at each step. It downloads a sample video (mice.mp4) for you. If you would rather be shown than read, start there.

  • Desktop app: the full loop — new project, video, suggestions, skeleton, labeling, training, correcting predictions, retraining, and inference.
  • Browser: everything up to training — new project, video, suggestions, skeleton, and labeling one frame. It ends there, because models can't be trained in the browser.

1. Open the app

Go to app.sleap.ai. Nothing to install.

Launch SLEAP after installing it.


2. Open a project — or start one

If you already have a .slp file, drag it onto the window, or use File ▸ Open Project… (Cmd+O / Ctrl+O).

.pkg.slp files with embedded videos work too, and open with no further setup — the frames come out of the file itself.

To start fresh, use File ▸ New Project… (Cmd+N / Ctrl+N):

  1. Pick a Skeleton — one of the built-in templates (fly, mouse top-down, human, C. elegans) or Empty — define later.
  2. Add videos: drag them onto the dropzone or click it to browse. No video handy? Click Use sample video to download mice.mp4.
  3. Click Create Project.

Where projects live

In the desktop app, projects are ordinary files on disk and Save writes back in place. In the browser, it depends on the browser:

  • Chrome / Edge — Save writes back to the file you opened, and Save As… opens a save dialog.
  • Firefox / Safari — both download a new copy of the .slp.

See Saving & Recovery.


3. Move around the video

Action Shortcut
Next / previous frame Right / Left
Jump 10 frames Cmd+Right / Cmd+Left (Ctrl+Right / Ctrl+Left)
Jump 100 frames Cmd+Shift+Right / Cmd+Shift+Left (Ctrl+Shift+Right / Ctrl+Shift+Left)
Next / previous labeled frame Alt+Right / Alt+Left
Next / previous suggestion Space / Shift+Space
Go to frame… Cmd+J / Ctrl+J
First / last frame Home / End

The seekbar under the video marks labeled frames, shows track occupancy bars, and can plot a per-frame statistic behind them — instance count, point displacement, prediction score, and more. See Navigation.


4. Place an instance

  1. Press Cmd+I / Ctrl+I (or Labels ▸ Add Instance) to drop a new instance on the current frame.
  2. Drag nodes to their correct positions.
  3. Nodes you can't see should be marked non-visible rather than guessed — right-click the node ▸ Mark Node Non-Visible.
  4. Cmd+Z / Ctrl+Z undoes anything.

Have more than one animal in the frame? Hold Ctrl and drag a node of an existing instance to clone it (it's Ctrl on macOS too, not Cmd), or right-click ▸ Add Instance.

More in Labeling Instances.


5. Pick better frames to label

Labeling consecutive frames is mostly wasted effort — neighboring frames look almost identical. Open the Suggestions panel and generate a set of frames spread across the video instead:

  • Stride — evenly spaced, the sane default
  • Random — uniform random sample
  • Frame chunk — every frame in one range
  • Image features — decodes frames, clusters them, and picks a diverse set
  • Prediction score / Velocity / Max displacement — target frames a model already struggles with

Then move through them with Space / Shift+Space.

See Suggestions.


6. Save

Cmd+S / Ctrl+S saves back to .slp. The first save of a new project suggests the name labels.v001.slp. File ▸ Save As… writes a new file and suggests the next version (labels.v002.slp, …).

The app also keeps a background draft of unsaved work, so a crashed tab or a closed window doesn't cost you labels — you get a Restore unsaved work? prompt next time. It's a safety net, not a substitute for saving.


What next?

  • ✏️ Label a real dataset


    Skeletons, placement methods, and the editing workflow in depth.

    Your First Labels

  • 🧠 Train a model


    Set up the Python environment and run sleap-nn training from the app.

    Training

  • ⌨️ Learn the shortcuts


    Every keybinding, matching SLEAP's defaults.

    Keyboard Shortcuts