Types of captchas

Sites use the word captcha for several different tests. The useful split is what the visitor is asked to do, because that decides whether a reader of images can help.

Text and numbers

A text captcha is a picture of letters or digits, usually warped, crossed with lines, or drawn in an odd font. A numeric captcha is the same idea limited to digits, often a fixed length such as six. These are the pictures on many Indian portals. Numeric captchas and distorted text are this family.

Arithmetic

An arithmetic captcha prints a sum, such as 14 + 9, and wants the result. A Hindi-number captcha writes one side as a Devanagari number word, such as बयालीस − 6, and still wants the integer. Those are pictures of a sum, not a paragraph of Hindi. See arithmetic captchas and Hindi-number captchas.

Photo grids and checkboxes

A photo-grid captcha asks you to tap squares of traffic lights, buses, or crosswalks. A checkbox captcha shows "I am not a robot" and may never show a puzzle if the vendor's risk system is satisfied. Both are browser products. They watch the page, the mouse, and the network. There is often no captcha image to download. Checkbox and invisible captchas describes that family. This API does not solve them.

Audio

An audio captcha plays distorted speech so a person who cannot see the image can still pass. It is a sound file, not a picture of a speaker icon. Audio captchas explains the limit: this service reads images only.

What to send here

If the portal gives you a PNG, JPEG, or similar file of characters or a short sum, that file is the input. If the page only offers a checkbox, a tile grid, or a sound, there is no image of the kind these models read. The image versus checkbox note is the short version of that choice.