Local-first AI dataset creation tool for serious ML builders. No cloud uploads. Just results.
A focused toolkit for building high-quality vision datasets — designed for speed, precision, and full local control.
Define any schema — numbers, text, dropdowns, booleans, multi-select. Your dataset, your structure.
Browse and select folders using your OS native dialog. No drag-and-drop hacks or path guessing.
Canvas-based crop tool with drag handles, aspect ratio lock, zoom controls, and instant preview.
Everything runs on localhost. Your images never leave your machine. Zero cloud dependency.
Every save appends structured data to a JSONL file — ready for Gemma, LLaVA, or any vision model.
Drafts are autosaved every few seconds. Close and reopen — your progress is exactly where you left it.
From raw images to structured JSONL — streamlined into the simplest possible workflow.
Point to your source images and choose an output directory. One click with native OS dialogs.
Define your label fields — numbers, categories, booleans. Set validation rules and required fields.
Crop, label, and save each image. Data exports automatically to JSONL format as you work.
A minimal, distraction-free workspace that keeps you in the flow.
Start building datasets today. Upgrade when you need more power.
Try the tool with up to 100 images per dataset.
Save 16%. Unlimited images, unlimited fields, and advanced features.