AI Textract

Turn PDFs and photos into structured data — text, forms, tables and receipts, priced per page.

AI Textract is document understanding rather than plain OCR. Users upload a PDF or image and choose one of four extraction types: text for raw content, form for key-value pairs, table for tabular structures, and receipt for expense data. Each type produces genuinely different output, so the choice is made up front rather than by post-processing one generic dump of text.

Billing is unusually careful. PDF page count is measured locally before anything is submitted, so credits are checked against the real page count and a user is never charged for a job that was going to exceed their balance. Each of the four types has its own per-page price, and multipage PDFs are processed as asynchronous jobs rather than held open in a request. Admins set independent size caps for images and PDFs plus a maximum page count, and supply their own AWS credentials, region and S3 bucket — meaning uploaded documents are staged in storage you control, which matters when the files are invoices, contracts or ID documents. Results are listed in a table with view and delete.

Key features

• Four extraction types — raw text, form key-value pairs, table structures and receipt expense data.
• Per-page pricing — each of the four types has its own credit cost per page.
• Accurate pre-billing — PDF pages are counted locally before the job is submitted.
• Async multipage handling — long PDFs are processed as background jobs.
• Separate limits — independent size caps for images and PDFs plus a maximum page count.
• Your own bucket — documents are staged in an S3 bucket you configure and control.
• Format validation — JPEG, JPG, PNG and PDF only, checked before upload.
• Results table — every extraction stored with view and delete.

Powered by

Powered by Amazon Textract with Amazon S3 for staging. You supply your own AWS credentials, region and bucket, so documents stay in infrastructure you control and usage lands on your existing AWS bill.

Ideal for

• Bookkeepers and finance teams digitising receipts and invoices.
• Operations teams processing forms and applications at volume.
• Researchers extracting tables from scanned reports.
• Anyone whose workflow currently involves retyping a PDF.