Synthetic-data sandbox

From sheet to checked row.

Read a printed timesheet, check it beside its source, and download the records you reviewed.

1. Choose your synthetic sheets

One supported layout: FIRSTLINE DEMO V1, printed English labels, one worker/assignment/week per file. PDF, PNG or JPEG; 1–10 files per session, 5 MiB each, 20 MiB total. PDFs: exactly one page, unencrypted. Images: at most 12 megapixels / 6000 px per edge. Handwriting is unvalidated; unreadable values must be reviewed or excluded.

Choose files or drop them here

Choosing a batch replaces the current session. Sample batch includes a PDF, scan and six exception cases.

Choose files or try the sample batch.

Download samples, template and demo rules

Use the clean PDF as the template; keep the labels, change synthetic worker/hours, then save as PDF or a clear printed image. Every weekday and weekend day needs an explicit number, including zero. Week is a Monday in YYYY-MM-DD format. Total is optional; if present, we compare it with the sum of daily hours, using decimal hundredths. No breaks, overtime, pay rates or rounding rules are inferred.

clean-text.pdf clean-scan.png missing-id.pdf duplicate.pdf inconsistent.pdf unsupported.pdf unmatched.pdf ambiguous.pdf heldout-text.pdf heldout-scan.png unreadable.png two-pages.pdf invalid.pdf encrypted.pdf

Inspect source values and expected results

Synthetic references

Matching uses exact worker ID plus name, exact assignment ID plus client, and the supplied week. No fuzzy matching or ID inference. A name shared by multiple workers is ambiguous.

Workers
ID	Name
W-104	Alex Reed
W-108	Jamie Lane
W-112	Robin Moss
W-116	Sam Vale
W-120	Sam Vale

Assignments
ID	Client
A-10	Demo Orchard
A-20	Demo Harbour

2. Check the source and review

Every record needs your review. OCR uncertainty is heuristic, not a calibrated accuracy score. Correct a field or resolve each exception with a visible note. Editing values clears that record’s notes and invalidates reviews; duplicate decisions must be checked again when the batch changes.

Document records
File / page	Worker	Assignment / client	Week	Hours	Status

Source

Raw extracted text (before corrections)

3. Download your reviewed rows

Only reviewed, included records go into the reviewed CSV. Unresolved, excluded and not-yet-reviewed records go into the separate exceptions CSV. Source references are local filenames and page numbers, not permanent hosted document links.

No records yet.