Everything you need
Built to give you an honest number.
gramscale isn't a party trick that spits out a fake-precise digit. Every feature exists to make the estimate as accurate as the sensors and the photo allow, and to say so honestly when it isn't sure.
Point-and-shoot object scanning
Camera capture with on-device downscaling before every upload, longest edge to 768px, EXIF and GPS stripped. Works on food, jewelry, packages, furniture, decor, and tools.
Swappable vision model
GPT-4o or Gemini Flash identifies the object and material and returns a structured weight estimate with a low–high range, selected server-side so the estimate quality can improve without an app update.
On-device length & dimension capture
ARKit measures length, width, and height fully offline. LiDAR-equipped iPhones get mesh-accurate raycasting; other iPhones fall back to plane detection, clearly flagged as lower accuracy.
LiDAR volume for offline weight
On supported iPhones, a measured volume from the LiDAR mesh combined with a material you pick gives an instant weight estimate, no network required, no vision model call needed.
Bundled material density table
A material-to-density reference covering every category ships with the app and loads on first launch. It powers offline weight math and sanity-checks the vision model's online estimate, widening the range if they disagree.
Offline queue
Every scan saves to your device first. If a weight estimate needs the vision model and you're offline, it's queued and refined automatically the moment you reconnect, nothing is lost.
Batch weighing (Pro)
Scan up to 20 objects in one sequential session with a running total and count, or capture several objects in a single photo. Export the summary as an image or CSV.
Scan history
Every estimate is saved to your account and synced across devices, source-labeled as AI, LiDAR + material, or Pending. Tap any past scan to review the full breakdown.