How do I batch OCR multi-page PDFs through an API?

To batch OCR multi-page PDFs through an API, the usual pattern is:

1. Upload the PDF
2. Create an OCR job
3. Poll job status or receive a webhook
4. Download the extracted text / searchable PDF / JSON output
5. Repeat in bulk with rate limiting and retries

Common workflow

1) Upload files

If the API requires pre-uploading documents, send each PDF first.

Example:

curl -X POST https://api.example.com/v1/files \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.pdf"

Response might include a file_id:

{ "file_id": "file_123" }

2) Start OCR jobs in batch

Submit one job per PDF, or a batch request if supported.

Example:

curl -X POST https://api.example.com/v1/ocr/jobs \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "file_id": "file_123",
    "output_format": "json",
    "language": "eng",
    "pages": "all"
  }'

Response:

{ "job_id": "job_456", "status": "queued" }

3) Check job status

Poll until complete:

curl -X GET https://api.example.com/v1/ocr/jobs/job_456 \
  -H "Authorization: Bearer YOUR_API_KEY"

Possible response:

{
  "job_id": "job_456",
  "status": "completed",
  "result_url": "https://..."
}

4) Download results

Depending on the API, results may be:

• plain text
• JSON with page/line/word coordinates
• searchable PDF
• HOCR / ALTO XML

Example:

curl -L "https://api.example.com/results/job_456" -o output.json

---

Batch processing pattern

If you have many PDFs, use:

• async job submission
• concurrency limits
• retry with exponential backoff
• webhooks if available instead of polling

Example pseudo-code

import time
import requests
from concurrent.futures import ThreadPoolExecutor

API_KEY = "YOUR_API_KEY"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}

def submit_and_wait(path):
    with open(path, "rb") as f:
        r = requests.post(
            "https://api.example.com/v1/files",
            headers=HEADERS,
            files={"file": f}
        )
    file_id = r.json()["file_id"]

    r = requests.post(
        "https://api.example.com/v1/ocr/jobs",
        headers=HEADERS,
        json={"file_id": file_id, "output_format": "json", "pages": "all"}
    )
    job_id = r.json()["job_id"]

    while True:
        s = requests.get(f"https://api.example.com/v1/ocr/jobs/{job_id}", headers=HEADERS).json()
        if s["status"] in ["completed", "failed"]:
            return s
        time.sleep(2)

pdfs = ["a.pdf", "b.pdf", "c.pdf"]

with ThreadPoolExecutor(max_workers=4) as ex:
    results = list(ex.map(submit_and_wait, pdfs))

---

Things to watch for

• File size limits: large PDFs may need chunking or preprocessing.
• Page limits: some APIs limit pages per job.
• Timeouts: OCR can take time; use async jobs.
• Rate limits: batch with limited concurrency.
• Output format: choose JSON if you need structured extraction; searchable PDF if you need document archives.
• Language/model options: specify language(s) for better accuracy.
• Rotation/de-skew: some APIs have image cleanup options.

---

If you’re using a specific OCR API

The exact endpoints differ. If you tell me which provider you’re using, I can give you:

• the exact request format
• a working Python/Node/cURL example
• a batching script for multi-page PDFs