Load a handle photo, rotate if needed, pan/zoom, then run OCR. Boxes are colored by confidence: green ≥0.9, yellow ≥0.5, red below; gray dashed = drawn but not yet recognized; dark red dashed = recognition tried and found nothing. Each box shows just a number by default — hover or select it (or its row in the list) to see its full text.
| Hover a box | Reveal its full text (default view just shows a number) |
| Left-drag on empty canvas | Draw a new box |
| Click a box | Select it (click again on its body to deselect) |
| Drag a selected box's body | Move it |
| Drag a selected box's corner handle | Resize it (editing resets it to unrecognized) |
| Ctrl + left-drag | Pan |
| Two-finger scroll | Pan |
| Pinch | Zoom (anchored at cursor) |
| Delete / Backspace | Remove selected box |
| Hover near a box (or select it), click the × floating above it | Delete that box immediately |
| Click a result in the list | Select/deselect its box |
| 🔍 / × next to a list row | Pan/zoom to that box / delete it directly |
| ⧉ / ∅ / 🗑 above the list | Prune overlapping boxes / prune empty boxes / clear all boxes (keeps the photo) |
| "Clear ▾" menu | Drop image (leaves the batch) / Clear batch (also deletes its saved boxes) / Clear all (deletes every saved box, ever) |
| Finish batch | Empty the batch for now; every image's boxes stay saved and reattach if you re-select the same files |