Azure OpenAI Service
An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
3,132 questions
Clarification on DALL·E 3 Image Generation Safety Rejection
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(217.60000000000002, 34%, 31%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMF%3C/text%3E%3C/svg%3E)
Magdi Fayed
0
Reputation points
I recently attempted to generate an image using DALL·E 3 on Azure with the following description:
"A 35-year-old man wearing a blue t-shirt that says 'I Love Dubai,' wearing short pin pants, holding his newborn daughter in a delivery room where his wife is resting on the hospital bed, smiling at both of them. There's a beautiful sunset visible from the window, and the atmosphere is very cheerful."
The task failed due to the safety system, and I'm unsure why the content was rejected, as the description seems appropriate.
In July and August 2024, I was able to generate images based on similar descriptions without any issues. Have there been recent changes to the content safety filters or policies that might be causing this problem? It seems that any description involving a newborn, baby, or child is now being rejected.
Is there a way to understand the reason for the rejection or to adjust the content safety settings to allow for such descriptions?
Thank you for your assistance.
{count} votes
-
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(281.6, 88%, 37%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ESA%3C/text%3E%3C/svg%3E)
Saideep Anchuri 75 Reputation points Microsoft Vendor2024-09-26T17:20:29.7166667+00:00 Hi Magdi Fayed,
Welcome to Microsoft Q&A Forum, thank you for posting your query here!
The error you're facing could be related to content moderation policies in Azure's DALL·E 3 or general image generation restrictions. Azure's DALL·E might reject descriptions involving specific sensitive scenarios like hospital settings or newborns.
To resolve this, try adjusting the prompt by:
- Avoiding sensitive details like "delivery room" or "newborn."
- Focus on the general atmosphere or less specific elements. For instance, describe the cheerful mood, family setting, or location, and avoid specific references to sensitive situations.
This should help in successfully generating the image.
Do let us know if that helps or have any further queries.
Thank You.
-
Saideep Anchuri 75 Reputation points Microsoft Vendor
2024-09-27T14:59:27.5666667+00:00 -
Magdi Fayed 0 Reputation points
2024-09-30T10:23:18.05+00:00 Hi Saideep, We have removed the newborn and delivery room references from the prompt, but the request is still being rejected. Could this be due to the severity level of the content safety filter? Should we consider applying our customized content safety filter to address this issue?
-
Magdi Fayed 0 Reputation points
2024-09-30T10:23:56.31+00:00 We have removed the newborn and delivery room references from the prompt, but the request is still being rejected. Could this be due to the severity level of the content safety filter? Should we consider applying our customized content safety filter to address this issue?
-
Magdi Fayed 0 Reputation points
2024-09-30T10:27:58.5966667+00:00 We have removed the newborn and delivery room references from the prompt, but the request is still being rejected. Could this be due to the severity level of the content safety filter? Should we consider applying our customized content safety filter to address this issue?
-
Magdi Fayed 0 Reputation points
2024-09-30T10:31:06.76+00:00 We have removed the newborn and delivery room references from the prompt, but the request is still being rejected. Could this be due to the severity level of the content safety filter? Should we consider applying our customized content safety filter to address this issue?
-
Magdi Fayed 0 Reputation points
2024-09-30T10:36:18.9066667+00:00 We have removed the newborn and delivery room references from the prompt, but the request is still being rejected. Could this be due to the severity level of the content safety filter? Should we consider applying our customized content safety filter to address this issue?
-
Saideep Anchuri 75 Reputation points Microsoft Vendor
2024-09-30T18:36:04.33+00:00 Hi Magdi Fayed,
Sorry for the trouble you are facing, for content safety filter, please refer to:
Still if you face any issue, please reach out to the Azure support.
If you have any further queries, do let us know.
Thank You.
-
Magdi Fayed 0 Reputation points
2024-10-07T10:43:33.53+00:00 Could we gain access to the word patterns/library that will be filtered out by Azure Content Safety? Additionally, is there a way to modify or paraphrase the user's input in the prompt to ensure it complies with the safety requirements?
-
Saideep Anchuri 75 Reputation points Microsoft Vendor
2024-10-07T17:44:36.1066667+00:00 Hi Magdi Fayed,
Azure AI Content Safety offers tools for moderating text and images, but it doesn't grant users access to the exact word patterns or libraries it uses for filtering. To modify or paraphrase the user's input in the prompt to ensure it complies with safety requirements. Here are a few approaches like paraphrasing and rewording, etc.
Thank You.
Sign in to comment
Sign in to answer