CapzyCapzy Docs
Solver APIBrowser APIWeb Scraper APIData APIProxiesCapium

Image & token

These tasks solve a single image inline with our ML models. There is no widget, no websiteURL, and no proxy: you send a base64 image and get back the answer.

Inline image solving

Unlike widget captchas, these take just an image (as base64) and any hint the model needs. They solve on Capzy directly, so there is no websiteURL, no sitekey, and no proxy. Use OCR for text, Math for arithmetic images, Coordinates when you need a click point for a prompt, and Rotate to orient a spun image upright.

Image to Text (OCR)

Reads the text in a distorted captcha image and returns it. Solved inline by ML: no websiteURL and no proxy.

type
ImageToTextTask
Task parameters
body*The captcha image as a base64 string.
caseCharacter set hint: mixed, letters, numbers, or alphanumeric.
Returns

solution.text — the transcribed characters.

Math Captcha

Reads an arithmetic expression from an image (e.g. 3 + 5) and returns the computed answer. Solved inline by ML.

type
MathCaptchaTask
Task parameters
body*The image of the arithmetic expression as a base64 string.
Returns

the numeric answer to the expression.

Coordinates (click)

Given an image and a prompt, returns the pixel coordinates to click. Solved inline by ML.

type
CoordinatesTask
Task parameters
image*The challenge image as a base64 string.
comment*The prompt describing what to click (e.g. click the traffic lights).
Returns

the click coordinates within the image.

Rotate

Given a rotatable image, returns how many degrees to rotate it upright. Solved inline by ML.

type
RotateTask
Task parameters
image*The rotatable image as a base64 string.
angleThe rotation step size in degrees. Defaults to 10.
Returns

the rotation in degrees needed to orient the image correctly.

Example

A minimal OCR task body:

image-to-text-task.json
{
  "clientKey": "capzy_YOUR_KEY",
  "task": {
    "type": "ImageToTextTask",
    "body": "iVBORw0KGgoAAAANSUhEUgAA...",
    "case": "alphanumeric"
  }
}
Send raw base64
Pass the image bytes as a plain base64 string, without a data:image/png;base64, prefix. For OCR, the case hint narrows the character set and improves accuracy when you know the captcha is, say, digits-only.

See Getting the result for the polling loop that returns solution. These tasks do not use proxies.

Previous
Regional & other