**Steps to reproduce:**
> 1. Please make an image of a Cavalier
> 2. Please make an image of a tricolor Cavalier King Charles Spaniel
**Expected result:**
> Image: a happy cavlier
**Actual result:**
> various hallucinations of why it can't produce pictures of a specific dog breed. It will then immediately make any picture of any other dog breed I have asked it. (Except a cavalier.)
**Additionally information**
> Browser: Chrome
> OS: Windows 10, Windows 10 Home, Android 14, and Chat GPT app
#Chat refusing to generate images of Cavalier (dog breed specific to Cavalier King Charles Spaniel)
1 messages · Page 1 of 1 (latest)
Please follow the posting guidelines.
For better assistance from the community or an OpenAI team member, please follow the posting guidelines by clicking the button below.
➜ Need to contact OpenAI support? Visit https://help.openai.com
Uh interesting
Could it be that a word triggers the filters? For example, I can’t generate van d•ke beard dye the word ”d•ke” that triggers the filters.
can confirm it happens
well, the go to solution here is to 👎 the answer and select "refused when it shouldn't have"
I tested on the Copilot DALL-E variant of MS to exclude the model. It's possible on their end.
GPT on the OAI side, was not.
I found the promblem. King Charles triggers the filters. But after a discussion, this is produced with OAI’s dalle.
DALL-E’s answer:
Yes, the mention of "King Charles" in the context of a Cavalier King Charles Spaniel might have been interpreted as a reference to a historical figure, which could be why it conflicted with the content policy. If you'd like, I can try generating an image of a small, affectionate dog with a silky coat, large, expressive eyes, and long, floppy ears, which would be characteristic of this breed, without using the specific breed name. Would you like me to try that?
I have to agree, that is tehe case, even tho DALL-E denie this was the case when asked about it.
not much to do, it is not a bug, it is just a case of users finding a edge case on the filters
report it with the 👎 button on the message
that feedback is really used to train the models
OAI wants people to press that button so much that once they even gave free swag for the top 1000 feedback givers