Document Types & File Requirements

Taggun processes receipts and invoices submitted as photos, PDFs, and digital files. This page covers what the API accepts and how common edge cases behave — it is the canonical reference for file and document questions.

Accepted File Formats & Limits

The 20 MiB limit applies to the complete HTTP request body. Multipart framing counts toward it, so a multipart file must be smaller than 20 MiB. For encoded endpoints, JSON and base64 expansion count toward it, so the decoded source file must be smaller again. An encoded JSON body of 20,971,520 bytes reaches application validation, while a body one byte larger returns 413 Request Entity Too Large. This body-size observation does not establish the maximum document Taggun can fetch from a URL.

Keep files as small as practical while preserving legibility. For request rate limits, see Usage Limits.

Image Size, Resolution & Format

File size, resolution, and format matter most because they affect readability and upload speed:

• Aim for readable, not tiny. A compressed photo around 1 MB is often a practical sweet spot for mobile uploads, but it is not a guarantee of accuracy and should not be treated as a hard target.
• Legibility is the real minimum. The receipt should fill most of the frame, and the smallest text you need extracted should still be readable after compression. There is no useful DPI target for phone photos.
• Use the file you have. JPEG, PNG, GIF, HEIF, HEIC, and PDF are all supported. PNG can preserve more detail in some cases, but converting a JPEG to PNG will not improve OCR accuracy.
• Avoid oversized images. Taggun automatically resizes images over 10,000 pixels in any dimension, but resizing adds processing time. Keep files as small as practical while still readable and keep the complete direct-upload request body within 20 MiB.

Document Types

Taggun attempts to extract common transaction fields—such as totals, tax, date, merchant, and line items—from supported receipt- and invoice-style documents, including:

• Receipts — point-of-sale receipts, printed or digital.
• Tax invoices — formal invoices carrying tax detail and tax IDs.
• Other business documents — Taggun’s documented extraction scope covers receipts and invoices. Other business documents may yield receipt- or invoice-like fields, but support, classification, field coverage, and extraction completeness are not part of the public contract unless the document type is explicitly documented. Test representative files before relying on an undocumented document type in production.

Document-type handling is endpoint- and country-specific. The documented schemas expose classification fields for Brazil, Chile, and Colombia.

The public extraction schema does not define a generic documentType field. Outside the documented country envelopes, do not infer classification from the general extraction response.

Digital vs Photographed Receipts

Both are supported. Submit digital receipts (emailed PDFs, screenshots) and photographed paper receipts to the same endpoints. If your workflow treats them differently — for example, a campaign that only accepts photographed receipts — Digital Receipt Detection classifies each submission.

Handwritten Receipts

Taggun processes both printed and handwritten receipts and invoices. Accuracy varies with the legibility of the handwriting — clear, well-lit images give the best results. To flag handwritten submissions (a common fraud signal in promotions), see Handwritten Receipt Detection.

Long Receipts & Multi-Image Submissions

Each API request accepts one file. That file can be one image or a PDF of up to five pages. The API does not accept an array of separate images in one request. For a receipt that cannot fit in one supported file, submit overlapping images as separate requests and combine the results in your application:

1. Split the receipt into multiple photos with some overlap between images.
2. Submit each image as its own request.
3. Combine the results on your end.

The overlap helps you stitch results reliably. Very long receipts also take longer to process — see Performance & Response Times.

Getting the Best Results

• Clear, well-lit, flat images with all edges visible extract best.
• Keep files as small as practical while readable — smaller files process faster, but over-compressing text hurts accuracy.
• If you're building a consumer capture flow, see End-User Capture Guidance for UX recommendations.

Need something not covered here? Contact [email protected] — see Contacting Support.