Free on iPad · No code · No cloud
Train where
you want.
Train image classification machine learning models right on your iPad. Import your images, press train, and export a ready-to-use .mlmodel — no code and no cloud.
Train image classification machine learning models right on your iPad. Import your images, press train, and export a ready-to-use .mlmodel — no code and no cloud.
Why MagicaLCore
Training a model usually means a Mac, a script and a cloud account. MagicaLCore puts the whole process on the iPad you already carry.
Folders in, model out. Every setting is a toggle, a slider or a stepper — with an explanation one tap away.
Training runs on your iPad. Your images stay on the device — nothing is uploaded to train a model.
Export a ready-to-use .mlmodel and drop it into an Xcode or Swift Playgrounds project.
Everything you need
Everything you need to build, check and ship an image classifier — in one app.
Pick one folder with a subfolder per class. MagicaLCore reads the labels from the folder names.
Name your project, add author and description, and set up training from a visual panel.
Your model trains on the iPad itself. No upload, no queue, no cloud account.
Iterations, learning rate, dropout, early stopping tolerance and iterations without improvement — each with an explanation.
Training and validation accuracy, loss, progress and memory, with live accuracy and loss charts.
Point the camera at something and see what your model predicts, in real time.
Import a folder of new images and check the predictions one by one.
Save a ready-to-use .mlmodel for Xcode and Swift Playgrounds.
Step-by-step guides inside the app, from your first folder to your first model.
Every model you train is kept with its author, date and description — ready to test or export again.
Put your images in one main folder with a subfolder for every class — Monstera, Ficus, Pothos… Import it from Files and MagicaLCore takes the labels from the folder names.
Go with the defaults or fine-tune the training. Tap the info button next to any setting and a short explanation tells you what it does.
Follow the training as it happens. Accuracy and loss for training and validation, iteration by iteration, with elapsed and remaining time.
Check your model straight away. Import a folder of new images and tap through the predictions, or switch to the camera and test it live.
Each model you train is saved to your History with its author, date and description. Swipe to test it again, export it or delete it.
MagicaLCore exports a ready-to-use .mlmodel. Save it to Files and add it to your app in Xcode or Swift Playgrounds.
New to machine learning? The built-in tutorials walk you through the app step by step — how to organise your images, train, test and find your models in the History.
The app
Designed and built natively for iPadOS. No web views, no cross-platform shortcuts — just your images, your iPad and your model.
On your iPad
Training happens locally, on the iPad in your hands. Your images stay on the device and are never uploaded to train. The models you save live in the app and, if you use iCloud, sync only through your own iCloud account.
Pricing
MagicaLCore is free with ads. Pro removes the ads and unlocks the advanced training settings and larger datasets.
MagicaLCore Pro, billed monthly. Cancel any time from your App Store settings.
MagicaLCore Pro, billed once a year.
Pay once and keep Pro for good. No subscription.
Q&A
