01 · Quickstart
First labels in five minutes
From empty account to a labeled, exportable dataset. Pricing is arranged with our team rather than a checkout, so start with the footage and we will size the work with you.
5 min · updated Jul 2026
Create your workspace
Sign in at app.lexannotate.ai. Your workspace holds projects; a project holds footage, labels, and everything Lexi learns about your world. Create one project per camera domain, one for rooftop surveys, one for line inspections, rather than one giant bucket.
tip · Name projects after the operation, not the date. 'Pipeline patrol' ages better than 'March footage'.
Upload footage
Drop in video or images. Video is split into frames automatically; high-resolution sources are preserved at full fidelity, so upload the original file, not a compressed export. A few minutes of representative footage beats an hour of the same shot.
Tell Lexi what matters
In the project chat, describe what you want found in plain language: 'Detect the poles, insulators, and any rust or damaged hardware on the line.' Lexi turns your sentence into classes and pre-labels every frame with tracked boxes.
tip · Be concrete about the failure you care about. 'Rust on the pin' outperforms 'anything wrong'.
Our team verifies
Pre-labels do not ship as they are. Our team walks them frame by frame: confirm what is right, fix what is close, reject what is wrong. Where we are not certain about a frame we flag it instead of guessing, so what comes back to you is verified work plus a short queue the console calls 'Awaiting your review'. Open those frames if the call is yours to make, or leave them with us.
Export
Export the dataset as COCO JSON from the project menu, boxes, classes, and image references, ready for any training pipeline. Or keep it in LexData and let the loop train on it directly.
tip · Exports carry verified labels. If 'Awaiting your review' still has frames in it, settle those first.
Fastest way to learn it is to run it.