Model files

The library doesn’t ship a model. You bundle one with your app and pass its file name to usePoseDetection or PoseDetectionOnImage.

Pick a model

Google publishes three pose landmarker models. They all output the same 33 landmarks; the bigger ones are more accurate and slower.

Model	Download	Best for
Lite	pose_landmarker_lite.task	Live camera on most phones. Start here.
Full	pose_landmarker_full.task	Better accuracy on recent devices
Heavy	pose_landmarker_heavy.task	Static images, or when accuracy matters more than frame rate

Put the file in a folder in your project, for example assets/models/pose_landmarker_lite.task, and point the config plugin at that folder:

{
  "expo": {
    "plugins": [
      ["react-native-mediapipe-posedetection", { "assetsPaths": ["./assets/models/"] }]
    ]
  }
}

On npx expo prebuild the plugin copies every file in that folder to:

Run npx expo prebuild --clean again whenever you add or change a model. See Expo config plugin for all options.

Android: copy the file into android/app/src/main/assets/ (create the folder if it doesn’t exist).

iOS:

In Xcode, drag the .task file into your app’s project navigator.
Tick Copy items if needed and make sure your app target is checked.
Under the target’s Build Phases → Copy Bundle Resources, confirm the file is listed.

Pass just the file name. The native side looks it up in the app bundle (iOS) or assets folder (Android):

usePoseDetection(callbacks, RunningMode.LIVE_STREAM, 'pose_landmarker_lite.task');