Distorted text captchas
2026-10-09 · Captcha basics
The classic captcha is warped letters or digits in a picture. Lines, colour, and odd fonts are part of the design.
The captcha most people picture is a short string drawn badly on purpose. Letters lean, digits touch, a line crosses the middle, and the background is speckled or coloured. AltaVista's filter around 1997 and the early Carnegie Mellon tests used this shape. It is still the default on a large number of portals.
The server stores the string it drew, shows the picture, and compares what you type. Refreshing the picture throws the old string away. A correct reading of an expired picture still fails, because the server has forgotten it.
Designers pick characters that look alike: O and 0, l and 1, S and 5. A reader has to know whether the field is letters, digits, or both. That is why a numeric code and a mixed word are different requests. Letter case and look-alikes is the product side of that choice.
Noise is not damage. The lines and the warp are the captcha. Cleaning the file in a graphics tool before you send it can erase the strokes a model expects. Send the original bytes. Noise, colour, and fonts says the same thing for this API.
Text captchas are the family Captcha OCR's text engine is for, including digit codes when you set numeric. A sum is the arithmetic engine. A photo grid is not a text captcha, even if the photos have letters in the background. Crop to the captcha image alone.
Ready to call it? Start with the quickstart, the API guide, or pricing.