Yeah also some people are deaf and literally cannot hear it.
I know some people like to play games in public places to pass the time with their volume off, it would be quite annoying to have to announce to the entire library: “ONE…NINE…FOUR…SIX…” as you enter in the captcha.
Audio is very hard to reproduce without making it plainly obvious what IDs were used to create it. I don’t think audio captchas would work.
Image captchas are a thought, and at best you wouldnt even need to design your own system. Just sign up for a good paid captcha service such as the one roblox uses, capture the images, render them to images, capture the caption, send them to the client, render them, make the player solve the captcha, send the answer to the server then the captcha, captcha success, BOOM.
I have recently been trying to make a captcha and this is one of the best solutions I came up with.
A server-side script would generate a captcha by using squares and pre-saved data on how to create letters. Randomness will be applied to make it really hard to create a list of possible combinations. Each square is an image with some random faint letters in it (The example below is just overlayed but it would be more inside) which makes it harder for AI, along with random squares in the background. With just an image, AI seems to not be able to decode what it says during my testing (Gemini AI & Google Image Search Text).
Of course hackers could come and separate the squares from the background and turn them all into just normal squares, then use some sort of AI to try to find the text, but out of the other solutions this is easier to render rather then a grid and somewhat simple to create, along with being much more secure then other options.
