Will any AI image model be able to generate correct images of a Colemak keyboard, by the end of 2024?
Dec 30

(harder version of /firstuserhere/will-any-ai-image-model-generate-co ?)

Must be to a single prompt, not an iterative prompting process or a conversation.

Must work for at least 10% of images generated.

As long as it can generate the A-Z keys in the correct order, an image will count as correct. Other keys are irrelevant.

The prompt must not tell the model the correct layout.

This is a Colemak keyboard layout, for reference:

Get Ṁ600 play money
Sort by:
predicts YES

I kinda think that, in addition to all the letters, the semicolon should be required to be in the right position, because it (unlike all other punctuation marks or the bizarre mapping of Tab to a second Backspace key) is located in the position of a letter (P) in the QWERTY layout.

predicts NO

There should be a market on whether anyone can get even one image with QWFP as the first four keys. Midjourney only, because midjourney generations can be shared with the prompt on the midjourney website and therefore can be verified.

More related questions