Image OCR: Bulgarian, Cyrillic, Greek & Latin avatar

Image OCR: Bulgarian, Cyrillic, Greek & Latin

Pricing

from $2.00 / 1,000 image with recognised texts

Go to Apify Store
Image OCR: Bulgarian, Cyrillic, Greek & Latin

Image OCR: Bulgarian, Cyrillic, Greek & Latin

Image to text: extract text from images and phone photos with OCR that runs inside the Actor. Tilted, sideways and upside-down pictures are straightened automatically. Bulgarian, Russian, Ukrainian, Serbian, Greek, English and more. No external API.

Pricing

from $2.00 / 1,000 image with recognised texts

Rating

0.0

(0)

Developer

Maine Wayne

Maine Wayne

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

4 days ago

Last modified

Features

Limited permissions

Share

Image OCR: Bulgarian, Cyrillic, Greek & Latin Text Extraction

Image to text: extract text from images (PNG, JPEG, GIF, BMP, TIFF, WebP) with an OCR engine that runs inside the Actor: no Google/Azure/AWS API, no API key, nothing is sent to a third party. Strong support for Bulgarian and other Cyrillic languages (Russian, Ukrainian, Serbian, Macedonian), plus Greek, English, Romanian, Croatian, German, French, Spanish and Italian.

Typical uses: digitising scanned letters, screenshots, receipts and invoices; reading text from photos for search; preparing images for AI agents and RAG pipelines; extracting Bulgarian text that general-purpose OCR tools often get wrong.

Input

FieldDescription
urlsPublic http(s) links to images (up to 100): PNG, JPEG, WebP, GIF, BMP, TIFF. robots.txt is respected. HEIC photos are not supported: export them as JPEG.
inlineImagesSmall images as base64 [{"filename": "a.png", "contentBase64": "..."}] (max 20, 10 MB total).
languages1 to 3 languages: eng, bul, ell, rus, ron, srp, srp_latn, hrv, deu, fra, spa, ita, ukr, mkd.
pageSegmentationModeauto, singleBlock, singleLine or sparse.
includeLines / includeWordsAdd lines/words with confidence and bounding boxes.
maxFileSizeMbSkip larger files (default 15, max 30).

Example input

{
"urls": ["https://example.com/scan.png"],
"languages": ["bul", "eng"],
"includeLines": true
}

Output (one item per image)

{
"source": "https://example.com/scan.png",
"status": "ok",
"format": "png",
"width": 900,
"height": 360,
"languages": ["bul", "eng"],
"text": "Столицата на България е София.\nТази поща е написана на кирилица: 1234 лева.",
"confidence": 95,
"wordCount": 12,
"preprocessing": { "rotationDegrees": 0, "scale": 1 },
"warnings": [],
"lines": [{ "text": "Столицата на България е София.", "confidence": 96, "bbox": { "x0": 30, "y0": 30, "x1": 700, "y1": 75 } }]
}

status is ok, no-text (nothing recognised) or error (see error). Only ok results are billed.

Ready-made examples

Saved examples you can open, run and copy:

Pricing

Pay per event: a small fee for each image with recognised text (image-processed), plus Apify's standard start fee. Images with no text and failed downloads are free. Set Max cost per run to cap spending; the Actor never processes or charges more than your limit allows. See the pricing panel on this page for current prices.

Tips for accurate results

  • Sharp images with text at least ~12 pixels tall work best (smaller text is enlarged automatically, blur cannot be repaired). The result carries a warning when confidence is low.
  • Pick the right language. Selecting bul for Bulgarian text is much better than eng. Combine languages only when both really appear (for example bul + eng).
  • Phone photos are fine. Pictures are straightened (tilt up to about 20 degrees), turned upright when they lie on their side or stand upside down, and enlarged when the text is small. preprocessing in the result says what was done; line and word boxes refer to the straightened picture.
  • Use singleLine for one line of text (captions, labels) and sparse for scattered words.
  • High-contrast, flat scans work best. Handwriting, stylised fonts, heavy backgrounds and curved text are weak spots of every OCR engine.

Limits

  • Up to 25 megapixels and 30 MB per image, 100 images per run. One image is stopped after 90 seconds.
  • Text recognition only: no translation, no handwriting model, no table or form understanding. PDFs are not accepted here: the PDF & DOCX to Markdown Actor reads PDFs, including scanned ones, with the same OCR engine.
  • Recognition is statistical: always spot-check important numbers and names.
  • You are responsible for having the right to process the images you submit.

FAQ

Is my image sent to Google or another service? No. The OCR engine (Tesseract compiled to WebAssembly) and the language data are bundled in the Actor.

How good is Bulgarian recognition? On clean, printed text it is typically very accurate; on noisy or low-resolution scans accuracy drops. Check confidence and warnings.

Can I read text from several languages in one image? Yes, select up to three languages, for example ["bul", "eng"].

Can an AI agent call it? Yes, through the Apify API or MCP server; use inlineImages for small images or urls for public ones.

Why was my URL refused? The site's robots.txt disallows automated access, the host is not public, or the file is too large. The error message says which.