One must be be an admin to do that.
I am a Discourse admin on another site, I lifted the URL from there and changed the site name for this forum. The image is also from that site which is why there are some emojis in that list that we do not see here.
Changing emoji list is not done often and even took me a few minutes last time I did it just to find the correct list, thus why I added the URL and since the name of that emoji is not so intuitive, why I also listed it.
Ah now I get it. And I probably read your post wrong, hence testing the link
I’m so used to use my phone, where I have all my emojis. But I wished more reaction options here (emojis). I saw that Discourse forum has more options.
… In case it not work for you.
You must use the model 4.5!
Up left you can select it under legacy. If you use the new one, it is multimodal model and generates the images directly. The link alone is not enough, it must be the old LLM too.
You could always start a new topic to start a Discourse poll to see what others would like.
I do see the poll option, no sure who has the ability to create them here.
A few problems I have learned about emojis:
@Daller oh yeah I got that. The link I referred to, was the one Eric posted. But I will test DALL•E and see the result
And @EricGT, yeah I totally get the problematic with having too many emojis as a reaction. I was more thinking about adding a couple of more, for example “
, 
, 
,
”…just to have a bit more reaction options. But it’s not the most important thing to add here on the forum, it’s just something I miss sometimes
(Sorry @Daller for this emoji reaction post)
99.99% of the users won’t even know Dall-E is deprecated and a new Image Generator 2.0 is available.
Of course only the users who use the image generator are meant, not all users. And that is only a small niche.
And before there was more activity in the forum around the image generators than now.
Yes, that is true. But keep in mind that in the last year, OpenAI has released a lot of important models that have nothing to do with image generation.
The problem is the background. Not that much the actual dragon. If you see the API generated one, you can understand the problem. Anyway, I tried again. It works better for the item in focus (i.e. dragon) but not for the background.
This question was recently posted as a new topic related to a problem using gpt-image-2
It was found that due to the design of my product, including bras and underwear. Always being prompted for violations. So, do you have any good suggestions
I tested this prompt
Professional e-commerce product photo of an adult female model wearing a matching bra and underwear set. Non-sexual commercial catalog photography, neutral standing pose, no suggestive expression, no erotic styling, no nudity, no transparent fabric, no provocative pose. Clean studio background, bright even lighting, product details clearly visible, accurate fabric texture, realistic fit, suitable for an online retail catalog.
Sharing a stable single-character anime prompt template that helps avoid multi-panel outputs and improves consistency
Hi,
I’ve been using ChatGPT Image 2.0 and built a general-purpose prompt template for generating stable single-character anime illustrations.
This helps avoid common issues:
unintended character sheets or multi-panel layouts
inconsistent proportions or anatomy
loss of personality in expressions and posing
It works as a simple core structure, and you can add extra elements (like animal ears, fantasy traits, etc.) without breaking stability.
Template
An anime-style illustration of [character type].
The character has [hair description], styled in [hairstyle].
The hair is natural with moderate volume, softly framing the face.
They have [eye description].
Their personality is [traits], expressed through [facial expressions and body language].
Body type: [body description].
They are wearing [outfit].
The pose is natural and expressive, reflecting their personality.
Composition: single character, no reference sheet, no multiple panels.
Use [lighting], [color palette], and a [overall atmosphere].
Focus on clean composition, strong silhouette, and natural visual flow.
Notes
Designed for single-character illustrations
Add ears, wings, accessories, etc. as needed
Avoid adding “reference sheet” or “multiple views” unless you want a model sheet
This structure has been much more stable for me compared to ad-hoc prompts.
Hope it helps 
The generator has multiple security layers. Both the text that is entered and a check of the finished image are used. So there can be several stages that filter the image. (At least simple trigger words are no longer used like in earlier generators.) But text meaning.
You often have to experiment a bit to describe the image in a way that doesn’t trigger directly. I would try to use terms related to fashion, clothing presentation, etc. And avoid: nudity, more then 1 person, interaction, etc. And also avoid brand names.
To find the trigger, use each phrase separately, or use a half-half search.
Before two trigger words were snow-white and black panther. So the security is a bit smarter now.
Here is a example from the past. “snow-white” and “black panther” wehre trigger words. So we used “pure white” and “black puma”. (But first we must spend time to find the triggers…) It is a bit the same on your case. Rephrase prompt parts witch could trigger a block, or a image result witch will be blocked.
A snow-white owl and a black panther are sitting peacefully next to each other. In the background, there is a beautiful landscape at sunset during the golden hour. Some flowers are growing around them. Photo style. (landscape format 1536x1024.) (don’t change the prompt, send it as it is.)
Welcome to the community!
We would be happy for you to share some of your images here:
oh and like maybe on the rock too—
Yeah, this happened for a couple of days, and sometimes I just thought it was extended use… but on the bright side, it hasn’t happened for a couple of days?
is yours alright now?
Hello everyone. I’ve been following this issue here in the forum for a few days now, because I’m affected by it as well. I don’t know how it is for everyone else at this point, but for me the problem still persists:
From a distance, the image looks atmospheric and well composed. But when zooming in, many areas break down into grainy, ornament-like micro-details. Stone surfaces look crumbly and artificial, edges are unclear, and vegetation, mud, and rocks merge into a noisy mass of texture.
This has been happening for several days, especially with fantasy, gothic, and landscape images. It does not just look like “a lot of detail”; it looks more like unwanted texture noise or overprocessing. As a thumbnail, the image can look good, but for larger display, printing, or projector use, it is completely unusable.
I create fantasy images that rely heavily on mood and atmosphere, often with a lot of fine-tuning. But when results like this appear even from a simple first prompt, it becomes a real problem.
How are you dealing with this?
Welcome to the forum!
Yes, I see clearly what you mean and that is an issue that many, (me included) have experienced. And I still sometimes experience quality issues with some images.
I just prompted this fantasy image in ChatGPT and this time, no unwanted noise.
Are you using ChatGPT or/and API to have image generated?
Please provide your prompt - I’ll see what I can do with it.
@Taragonn Yep, I have exactly the same problem like many many others. And it is technical in nature, you cannot prompt it away.
Anything that has natural or otherwise complex structures shows the pattern the strongest.
@LarisaHaster And unfortunately, I can recognize that pattern immediately, it is also in your image.
Quite honestly, the developers should decide whether they want to support a generator at all and then deliver a good result, or whether they simply want to let this AI line die honestly.
I think there will probably be some product sellers who can sell their perfumes and shoes. But for anything that has to do with creativity and imagination, the generator in its current state is not suitable, even though it has potential, which is especially disappointing right now.
… maybe tell him: Collection of GPT-image-generator 2.0 issues, bugs, and work-around tips (check first post) - #25 by polepole
For me, the image generator no longer makes sense, as long it delivers this quality.
It is as if you were generating good music, and then compressing the data with 64 kb/s, or JPEG with 20%. You simply never want to listen or watch it. I am still waiting to see whether they fix the problems. But it could be that they leave it like this. Unfortunately, then I am out. I will not use the generator like this. (I will be around the forum, but not make tests with this image-generator in the present state.)
I always advice: enjoy the few weeks and months when something is new, bevor it gets enshittifiied.
Disappointing!
It is curious that you are the one who went out of your way to create this topic. And your primary purpose seems to be a continuous rag on gpt-image-2. I understand that you are disappointed that your precious DALL-E 3 is gone - I am too. But people are doing amazing things with gpt-image-2 all over the world and it’s a shame that you cannot accept that.
