PDF OCR API: Run OCR from Your Code

What a PDF OCR API is for

• You get text you can search, index, or pass to the next step

• Scanned pages stop being “image only”

• You are not yet extracting named invoice fields (that is extract), and you are not chatting with the file

Practical rule: if the next step is Cmd-F in a downloaded file, the browser OCR tool may be enough. If the next step is your server, use the API.

Browser OCR vs PDF OCR API

Limits you should check first

Web app OCR (pricing page)

• Hobby: Free forever ($0), OCR 2 pages/file, max file size 10MB

• Pro: OCR 10 pages/file, max file size 50MB

• Ultimate: OCR 50 pages/file, max file size 50MB

• Enterprise: OCR 100 pages/file, max file size 100MB

API credits (developer docs)

• Parse Standard (OCR): 1 credit/page

• Parse Advanced (VLM): 2 credits/page

• Extra credits may apply if images need LLM analysis (llm enabled)

• Cached parse results with the same settings use 0 credits

• Free API plan: $0/month for 200 credits/month, no credit card required

• Paid API examples on that page: $49 / 3,000, $99 / 10,000, $249 / 30,000, $599 / 100,000 credits per month

Practical rule: pick the browser tool when you need a searchable file once. Pick the API when OCR must run without a human at the upload box.

How to call a PDF OCR API (Parse, standard)

1. Get an API key

2. POST to Parse with standard quality

• file or url (or docId when you already have a cached document id)

• quality: use standard for the OCR path (default). advanced uses VLM instead

• lang_list: languages for OCR when using standard (default ["en"])

• llm: optional image LLM processing (default false)

import requests

url = "https://pdf.ai/api/v2/parse"
headers = {"X-API-Key": "YOUR_API_KEY"}

with open("/path/to/document.pdf", "rb") as f:
    files = {"file": f}
    data = {
        "quality": "standard",
        "lang_list": '["en"]'
    }
    response = requests.post(url, headers=headers, files=files, data=data)

print(response.json())

3. Check the output before you trust it

• Search for a word from the first page and a later page you care about

• Check tokens that are easy to mangle: names, dates, amounts, IDs (0/O, 1/l, 5/S)

• If the page was a bad scan, fix the scan (or split a file that is too long for your path) before you blame the API

Practical rule: a non-empty markdown string is not a verified page. Read the lines you will use.

When the API path is the wrong tool

• You only need one searchable PDF. Use https://pdf.ai/tools/ocr-pdf, then prove search works. See OCR a Scanned PDF.

• You need named fields (invoice number, total, vendor). That is extract, not OCR. Start from the Extract pillar draft Extract Data from PDF: Get Fields and Tables Out and the live extract page.

• You need layout as structured blocks for RAG. Still Parse, but you are using the full parse result, not “make this file searchable” alone. The playground is https://pdf.ai/parser.

• You wanted a generic “PDF API” overview. That is a different topic. This page stays on OCR via API.

A short action plan

1. Decide: one-off searchable file, or OCR inside your code.

2. For one-off: open https://pdf.ai/tools/ocr-pdf, check pricing page/size caps, download, then prove search.

3. For code: create an API key, call POST https://pdf.ai/api/v2/parse with quality=standard, check credit usage.

4. Verify text from early and late pages before you wire the next step.

5. Keep the original PDF until the text you will use is correct.