VisionParse turns any document into structured data.

VisionParse converts scans, PDFs and photos into text and JSON with word-level positions. It reads 161 languages and reports a confidence score for every word. Free tools run in the browser without an account. Uploaded files are deleted within the hour on every plan, free and paid alike; extracted results are kept for 30 days so you can fetch them again. Documents are never used to train models.

  • No account needed
  • 161 OCR languages
  • Deleted within the hour
  • Never used to train models

Measured · Senithu Software Solutions

No file yetReady

Drop an image to read it

JPG, PNG, HEIC, TIFF, WEBP, BMP — up to 3 MB. Your file is read on our servers in Frankfurt and deleted within the hour.

5 files a day3 an hour3 MB a file3 pages a filedeleted after 1h

What VisionParse does, in numbers we measured.

  • VisionParse converts scanned documents, PDFs and phone photos into plain text and JSON carrying a position and a confidence score for every word.
  • The production server has 161 OCR languages installed, counted on the server on 31 July 2026.
  • A digitally generated PDF is read from its own text layer in 0.057 seconds, returning the exact characters rather than a recognised guess.
  • Recognising that same page as an image instead takes 0.508 seconds and misreads "AI" as "Al", which is why the text layer is always preferred.
  • Measured throughput is 7.58 digital pages a second and 3.13 scanned pages a second at six concurrent workers on eight cores.
  • The free tools run without an account at five conversions a day, three an hour, three pages and 3 MB a file.
  • Uploaded files are deleted within the hour on every plan, free and paid alike, and no document is ever used to train a model.

Convert a document right now.

Free, in the browser, without an account. 30 tools are live today; the rest publish as each is measured on its own format.

See all 30 tools and what is live

What it gets wrong.

Three things we measured on our own service. Published because a limit stated plainly is worth more than an accuracy percentage nobody can check.

  • Held up

    A digital PDF is read exactly, not recognised

    Anything exported from a word processor, invoicing system or CV builder already carries its characters. We read them directly: exact text, nine times faster than recognising a picture of the same page.

    0.508s recognised0.057s read

  • Known limit

    Capital I becomes lowercase l on recognised images

    When a page has no text layer and must be recognised from pixels, a capital I in a sans-serif face is indistinguishable from a lowercase l. On a real CV at 150 DPI this turned a subject heading into a different word.

    AI & Data ScienceAl & Data Science

  • Known limit

    Arabic returns the right letters in the wrong order

    Arabic script is recognised accurately character by character, but comes back in visual rather than logical order, so the words read backwards. Right-to-left output needs reordering before it is usable. Arabic-Indic numerals are also unreliable.

    فاتورة ضريبيةةيبيرض ةروتاف

How these were measured

Method: single-page samples, warm runs, 8 vCPU AMD EPYC / 23 GiB, Ubuntu 24.04. Small synthetic and real-document samples, not a full corpus. Throughput measured at six concurrent workers with the queue drained between runs.

Two calls, and the document is text.

Send a file, take the job id, read the result. Recognition never runs inside your request, so a forty-page PDF cannot hold a connection open until it times out.

The API needs a paid plan. The free tools above are web-only and need no key.

# 1. Send the document
curl -X POST https://api.visionparse.app/v1/ocr \
  -H "x-api-key: $VISIONPARSE_KEY" \
  -F "file=@invoice.pdf" \
  -F "languages=eng"

# 202 Accepted
{ "jobId": "c2022d05-1e92-4cc5-8321-a479d39ce552", "state": "queued" }

# 2. Read the result
curl https://api.visionparse.app/v1/jobs/c2022d05-1e92-4cc5-8321-a479d39ce552 \
  -H "x-api-key: $VISIONPARSE_KEY"

# 200 OK
{ "state": "completed",
  "result": { "text": "...", "confidence": 100, "pageCount": 1 } }

Free to read documents. Paid when they run your workflow.

The free tools stay free and never need an account. Plans add volume, the developer API and team features — and overage is one-time packs, never a surprise invoice.

Free

Web app

$0/month

100 OCR pages a month — or 5 files a day with no account at all

For a document now and then — free, and it stays free.

OCR only, web app only — no AI extraction, no API keys.

Starter

Developer API

$19/month

2,000 OCR pages a month

The API, for one developer or one workflow.

One seat.

Professional

Most chosen

$59/month

12,000 OCR pages a month

For a team putting documents through every day.

No audit log export or SLA — those are Business.

All plans and credit packs