Skip to main content

OCR Specialist

Extract text from images with Tesseract.

Overview​

  • Endpoint: ocr-public
  • Engine: Tesseract OCR (pytesseract)
  • Capability: image.ocr
  • Backend: specialist, CPU-only (vram_required_mb = 0)
  • Price: per request, USD — see the catalog for the current rate

When to use​

✅ Digitizing scanned documents, reading receipts/forms, pulling text out of screenshots and photos.

❌ Complex layout understanding, handwriting, or document Q&A — Tesseract does plain text extraction, not layout analysis. For a vision-language approach, try the multimodal qwen3-vl-8b.

Request / response​

Pass an image_url (the node downloads it and forwards the bytes) — or image_base64 — plus an optional Tesseract language code. You get back the extracted text and an average confidence.

from colabhive import ColabHive

client = ColabHive(api_key="hive_...", account_id="YOUR_ACCOUNT_ID")

result = client.endpoints.infer(
    endpoint_id="ocr-public",   # SDK resolves the name to a UUID
    input_data={"image_url": "https://example.com/receipt.png", "language": "eng"},
)
data = result["result"] if result.get("status") != "queued" \
    else client.endpoints.get_task(result["task_id"])
# data → {"text": "...", "language": "eng", "confidence": 91.3, "inference_time_ms": 120.0}
FieldTypeDescription
image_urlstringImage URL — fetched and base64-encoded by the platform
image_base64stringAlternative: send base64 image bytes directly
languagestringTesseract language code (default eng; e.g. spa, fra, deu)
configstringExtra Tesseract config flags (optional)

You can also upload a local file first with client.endpoints.upload_input(...) and pass the returned URL. The authoritative schema is GET /api/builder/v1/endpoints/{id}.

Tips​

  • Higher-resolution, high-contrast images read far better than small/blurry ones.
  • Set the right language code — the default is English (eng).
  • confidence is an average over detected words; low values flag a poor scan.

Next steps​

Authors: José Luis Minich, Maximiliano Lucius.