Free CAPTCHA solver & code reader

Add the image

Drag a file in, click to browse, paste straight from your clipboard, or hand it a public image URL. Base64 and data URLs work too.

It reads automatically

A convolutional recognition model runs on our server the moment the image lands. Typical images finish in well under a second.

Copy the text

The characters appear beside the preview with a copy button, the character count and how long the read took.

Great at

• Classic image CAPTCHAs — one line of warped letters and digits, roughly 4–8 characters
• Noise, speckle, strike-through lines, colour gradients and per-character rotation
• Short verification and one-time codes rendered as images
• Serial numbers, licence keys and reference codes on a plain background
• Digit-only strings such as 481920

Not built for

• Paragraphs, documents, receipts or book pages — use a document OCR engine instead
• Multi-line images; only a single line is read
• Spaces, which are always dropped — Order 12 returns order12
• reCAPTCHA and hCaptcha, which are behavioural and image-grid puzzles with no text to read
• Guaranteed letter case, and lookalikes like I/l or 0/O

Endpoint

POST https://ocr.aticmatic.com/ocr

# a file from disk
curl -X POST https://ocr.aticmatic.com/ocr \
  -F "file=@captcha.png"

# a public image URL
curl -X POST https://ocr.aticmatic.com/ocr \
  -F "url=https://example.com/captcha.png"

# base64 as JSON
curl -X POST https://ocr.aticmatic.com/ocr \
  -H "Content-Type: application/json" \
  -d '{"base64":"iVBORw0KGgoAAAANSUhEUg..."}'

import requests

URL = "https://ocr.aticmatic.com/ocr"

with open("captcha.png", "rb") as fh:
    r = requests.post(URL, files={"file": fh}, timeout=30)

r.raise_for_status()
print(r.json()["text"])

const body = new FormData();
body.append("file", fileInput.files[0]);

const res = await fetch("https://ocr.aticmatic.com/ocr", {
  method: "POST",
  body,
});

if (!res.ok) throw new Error((await res.json()).detail);
const { text } = await res.json();
console.log(text);

Request & response

Send exactly one of three inputs, as multipart form data or JSON:

• file — a multipart image upload
• url — a public HTTP or HTTPS image URL
• base64 — raw Base64 or a data URL, as a multipart field or a JSON key

{ "text": "7fk2p" }

Errors return the same shape as FastAPI's default, with a readable message:

{ "detail": "Image must be 8 MB or smaller." }

413 too large, 415 wrong content type, 422 not a readable image, 429 rate limited — retry after the seconds given in the Retry-After header. Images cap at 8 MB and fair use is 30 requests per minute per IP. Need more headroom or a private deployment? Talk to us.

Interactive docs

The full OpenAPI schema, with a browser console for trying requests, lives at /docs and /openapi.json.

Self-host it

The service is a single FastAPI file on top of ddddocr, with a Dockerfile in the repo. Run it on your own box in a couple of minutes.

Fair use

Please use it on CAPTCHAs you are entitled to solve — accessibility needs, your own forms, your own test suites and your own automation. Don't point it at other people's sites to bypass their protections.