Spellcheck in prompt inputs?

[nabileisa]

Hey , with the rising popularity of z-image and flux2-kelin, smaller models that are fast enough for most users that rely on natural language but require long descriptive prompts for their prompting and not tags , how hard would it be to get spellcheck working for the prompt inputs to avoid making mistakes with these long prompts that may result in bad output or or in wrong concept or image being generated?

[shadowlocked]

I only ever saw a spellcheck in Adobe Photoshop's prompt window - made a lot of sense. On the Mac, I'm pretty sure you can just plug into a native spellcheck service that's available to any application. And if I remember rightly (from hooking up Notepad++ with spellcheck on Windows), the Windows platform has something similar one could hook into.

[Acly]

I really don't think you need to worry about it. New models use LLM like Qwen2/3 as text encoder, they can deal with spelling mistakes. Text doesn't have to be perfect, they understand.

Afaik Qt (Krita's UI framework) doesn't have built-in spell check, so it would be a ton of work actually.

[nabileisa]

Understandable , I have thought as much from a quick google search before posting but since Python/Qt is not really my area of expertise I thought some one else might have other ideas , that's why I posed it as a question.

The reason I Asked is because English is native language so some times I tend to make some mistakes when typing in English that end up altering the meaning of a sentence even with LLM based text encoders but it's uncommon so it's bearable.