Fill from Images and Documents and Documents

How to use this docs page

This page is meant to answer one operational stage of the DullyPDF workflow well enough that you can run a controlled test without guessing. Read the sections below, validate the behavior against one representative document, and only then move to the next linked page.

That order matters because most setup failures come from mixing detection, mapping, fill validation, and sharing into one unstructured pass. A narrower review loop keeps troubleshooting faster and makes the template easier to trust once you save it for reuse.

What Fill from Images and Documents does

Fill from Images and Documents lets you upload photos or scanned documents (IDs, invoices, pay stubs, utility bills, medical records) and have DullyPDF extract matching information into your template fields automatically using OpenAI vision.

The pipeline sends your uploaded images alongside the template field schema (including nearby label text extracted from the PDF) to OpenAI. The model reads the uploaded documents, matches extracted data to your form fields by semantic meaning, and returns values with confidence scores.

Pipeline details

1. Click Fill from Images and Documents in the toolbar (requires named fields and an active session).
2. Click Upload to select one or more images or PDF documents. Multiple files can be uploaded at once.
3. Review the uploaded file list. Remove files with the x button if needed.
4. Check the credit cost in the footer: each image = 1 credit, each PDF document = 1 credit per 5 pages.
5. Click Send to run extraction. DullyPDF renders the template PDF to extract label context for each field, encodes your uploaded files, and calls OpenAI vision.
6. Review extracted fields: each shows the matched field name, the extracted value (editable), a confidence percentage, and a Reject button.
7. Edit values inline or reject fields you do not want applied.
8. Click Fill to write accepted values into your template fields. Existing values stay in place and are merged with the extracted information.
9. Or click Fill & Clear to wipe every field in the current PDF first and then apply only the extracted values. Use this when the template has prior data from a different record that should be fully replaced rather than merged.

What gets sent to OpenAI

• Your uploaded images and documents (encoded as base64).
• The template field schema: field names, types, and the label text printed next to each field on the PDF.
• A system prompt instructing the model to extract matching information.
• Your PDF template page images are not sent. Only the text labels near each field are included for context.
• Field values, row data, and respondent information are never sent.

Credit cost

• Each uploaded image (JPG, PNG, etc.) costs 1 credit.
• Each uploaded PDF document costs 1 credit per 5 pages (bucketed per document, rounded up).
• The dialog footer shows the estimated cost before you click Send.
• Credits are deducted from the same OpenAI credit pool used by Rename and Map operations.
• If the extraction fails, credits are refunded automatically.

Best practices

• Name and rename your fields before using Fill from Images and Documents. The AI matches by field name and label context.
• Upload clear, well-lit photos. Blurry or partially cropped documents reduce extraction accuracy.
• For multi-page documents like invoices or medical records, upload as PDF rather than photographing each page separately.
• Review confidence scores. High confidence (80%+) values are usually correct. Low confidence values should be verified.
• Use Reject to exclude fields you want to fill manually.
• Fill from Images and Documents works best for structured documents: IDs, invoices, tax forms, insurance cards, pay stubs, and similar.

Supported file types

• Images: JPG, JPEG, PNG, GIF, WebP, BMP, and other browser-supported image formats.
• Documents: PDF (rendered page-by-page for the AI model).
• Maximum file size: 20 MB per file. Maximum 10 files per extraction.