Skip to main content
0Sign in
All documentation

Supported formats and processing limits

Image/PDF formats, guest and account limits, byte targets, and unsupported document features.

Updated

This page is the reference: every number here is the limit the service actually enforces, not a marketing target. When a portal rejects your file and tells you why, come back here to see whether the ceiling is on your side or on ours.

Limits at a glance

Access levelFiles per jobCombined input sizePDF pagesPixelsDaily allowance
Guest, no account35,000,000 bytes2012,000,0005 jobs per browser session, and 30 per source network
Free account1025,000,000 bytes10040,000,00020 jobs per UTC day
Pro account10025,000,000 bytes10040,000,000100 jobs per UTC day

Read the table by column, not by row, because the three ceilings are enforced independently.

  • Files per job counts how many inputs you may hand to a single task. Merging eleven PDFs on a free account fails before any processing starts.
  • Combined input size is the sum of all inputs, not the largest one. Three 10 MB files are 30,000,000 bytes and exceed the 25,000,000-byte ceiling even though no individual file does.
  • PDF pages applies to the document you build as well as the one you supply. Merging two 60-page PDFs produces 120 pages, which exceeds the 100-page limit even though each input was legal.
  • Pixels is width × height of a decoded image. A 8000 × 5000 photo is 40,000,000 pixels, exactly at the ceiling. Very large phone panoramas and 100 MP camera files will exceed it.
  • Daily allowance resets at 00:00 UTC, not at your local midnight. If you are far from UTC, the reset will feel like it lands in the middle of your working day.

The guest row is not just a smaller version of the free row: it is also shared. The 5-job allowance follows your browser session, while the 30-job allowance is counted per source network, so a household, office or campus behind one address shares that pool.

Why these numbers exist

The 25,000,000-byte and 100-page ceilings come from the memory available to a single processing run. A PDF has to be parsed and re-emitted in one pass, and a 200 MB scan or a 500-page document would consume more memory than a job is allowed. The 40,000,000-pixel image ceiling is the same idea for raster data: decoding an image expands it far past its compressed size, and a 100 MP file expands to roughly 400 MB in memory before encoding.

On the guest side, the tighter numbers keep an anonymous, unauthenticated endpoint from being used as free bulk compute. They are deliberately generous enough for a person submitting an application and too small for a pipeline.

Format support

FormatReadWriteNotes
JPEGYesYesNo transparency. Converting to JPEG flattens transparent areas onto a solid background.
PNGYesYesLossless. The largest output of the common formats.
WebPYesYesSupports both lossy and transparency.
AVIFYesYesSmallest files, slowest to encode and least widely accepted by older portals.
HEIC / HEIFYesNoCommon on recent iPhones. Convert before uploading somewhere that expects JPEG.
Animated GIF / WebPYesNoProcessed as a single still frame. The animation is not preserved.
PDFYesYesSee the document-feature restrictions below.

Output choices depend on the operation, so check the format selector on the tool page rather than assuming every tool writes every format. If you are unsure what you actually have, inspect an image first — a .jpg extension proves nothing about the bytes inside the file.

PDF features that are rejected

Password-protected and encrypted PDFs, files with embedded attachments, and documents containing active content such as JavaScript or external actions are refused outright. There is no override. Export a plain PDF from the application that created it, if you have the rights to do so.

A PDF that carries a digital signature can be processed, but rewriting the file invalidates the signature. Keep the signed original and work on an unsigned copy. AttachReady cannot sign documents, and it cannot edit PDF text or repair a corrupt file — those operations do not exist here.

Size targets are measured in bytes

The single most common source of a failed submission is the difference between KB and KiB. A form that says "200 KB" may mean either of two different numbers:

LabelInterpreted asActual bytes
200 KB, decimal200 × 1000200,000
200 KiB, binary200 × 1024204,800
2 MB, decimal2 × 1000 × 10002,000,000
2 MiB, binary2 × 1024 × 10242,097,152

Our default target is 200,000 bytes, the stricter of the two readings. That is a policy choice, not rounding: aiming at the smaller number satisfies both interpretations. If the portal publishes an exact byte count, use that number instead, and leave a small margin — a result that lands at 199,999 bytes is fine, but a target you only just meet has no room for a re-encode.

What compression can and cannot do

Lossless cleanup alone rarely shrinks a scanned PDF by much, because the pages are already images. Real reductions come from recompressing those page images, which discards detail. After compression, open the result and zoom in on the smallest text. A file that passes the size check but is no longer readable has failed the real test. The scanned PDF guide walks through that trade-off.

Pixel dimensions and DPI are separate properties. Changing a DPI field in metadata does not add pixels and does not guarantee that a receiving site will accept the file.

Accounts, entitlements and API access

Public tool discovery does not require an API key; authenticated processing does. A failed attempt still consumes an allowance, so fix the obvious problem — a wrong file type, an oversized batch — before retrying, rather than clicking through the same error repeatedly. Tool-specific limits and aggregate checks can be stricter than the table above.

Purchasing a higher tier or credit packs is not currently enabled. A pricing page and the Pro row in this table describe what the plans will allow; they do not grant an entitlement today.