Back to Discussions
Bug Report

The prompt isn't working properly.

Mmini burger
Translated to English·

There might be changes related to prompt notation, and items displayed in the original language in my profile's creation history may not be created properly due to translation issues. On the main page, calling a normally functioning creation history, editing it, and then inserting different text may sometimes work. This is urgent, so sorry if the wording is odd.

2

9 replies

I feel like sometimes the prompts I input in Japanese aren't reflected properly 🤔

M
mini burgerOP·

It seems okay if it's just English text.
There might be an issue with the part that stores input.

たぬー·

In my profile, the prompts for my works are remaining in Japanese, which is causing completely unintended results. It seems there is an issue during the conversion process. Since Chappy seems to understand Japanese, this problem might not occur.

M
mini burgerOP·

It seems that V3 and Neta also recognize Japanese. For now, I've confirmed that entering a vertical hole causes abnormal behavior.

たかはしがに·

I feel like this is indeed not functioning properly in Japanese. When I copy and use a previously used prompt, it works, but if I add, delete, or modify that prompt, something completely different results.

For example, if I quote a prompt from the history of images created with prompts like "black hair, red eyes, sailor uniform," it works in Japanese, but if I add an arrangement like "black hair, red eyes, sailor uniform, sitting on a chair," it suddenly becomes blonde hair or blue eyes, showing that the prompt is not functioning at all. Probably a bug.

ねむ·

I am also encountering the bug where the image is not generated according to the prompt.
As far as I can confirm, the bug is occurring in Anim4gine, Animij, WAI, and V2 (maybe SDXL too?)
I discovered one thing: when I published a work that was not generated according to the prompt, I noticed that on the detail page of that work, the prompt is displayed in Japanese instead of English!

Also, this is trivial, but I am thinking of taking advantage of the current bug to create unexpectedly interesting works 😅

ねむ·

P.S.
I discovered that SDXL understands about half of the prompts. When generating images of people, words referring to the style or appearance of the picture are understood, but it seems that words about actions or backgrounds are not understood very well…

Leave a reply

Join the community to share your thoughts and ideas.

Sign in to join the discussion