Prompt
a tiled courtyard fountain, resting under soft daylight, with a flock of birds in the distance, comic book illustration
You don’t need a “prompt” to “modify” something for the AI to understand languages.
gpt-image-whatever is a multimodal language model, trained the same as gpt-4o was with a native (and unreleased for the longest time) ability to produce audio and imagery outputs. It can understand and produce in world languages, just as it can talk.
The only assurance needed is that text-to-output and image-to-output post-training then infers and extrapolates well in equal quality across languages.
It is powered by large language model AI intelligence.
Entity identification
Knowledge-based presentation
Interesting. Does OpenAI have published information on gpt-image-whatever’s multimodal capabilities? Or is this something that will be revealed someday when the SuperApp is released?
Enjoy a presentation in inscrutable AI image form.
https://openai.com/index/introducing-chatgpt-images-2-0/
Particularly
It can understand and produce in world languages, just as it can talk.
Well… kinda, but with smoke and mirrors. Multi-language understanding does not imply the process in which it is accomplished.
You may want to read this:
Edit: Over two years ago i started using OpenAI exclusively for document translations - 132 languages. Spent months testing. As the models progressed, the translations got better, especially for around 82 of the languages. But I found that I had to use system > developer > instruction prompts to improve translation quality.
we’re all clearly hallucinating
^.^
Yeah, the attached PDF is a head spinner, for sure. Gave me a headache…
Awhile back I got AI to create a one page short story about a young girl and a baby dragon - cute story. Then tested translations: I would translate to a language and then translate back to English to ascertain accuracy.
So, I translated the story to an African language (I forgot which one). But when I translated back to English, there was an error. The word “Dragon” was replaced by the word “Crocodile”. Why? Because that particular African culture had no idea what a friken dragon was 
Well that makes sense… it could have at least been a flying crocodile ^.^
A small tip for those using ChatGPT for image prompting:
On May 5, OpenAI improved memory in ChatGPT for Plus and Pro users.
ChatGPT can now better pull in relevant context from your past chats, saved memories, and, where available, files and connected Gmail app, to offer more tailored ideas, recommendations, and next steps. It is also faster at searching past conversations to find the right context, so you don’t have to repeat yourself as often.
(If you’re interested to read more about it, check this link: https://help.openai.com/en/articles/6825453-chatgpt-release-notes )
And I unintentionally tested this with a very simple image prompt, just writing: “generate an image that is beautiful, magical and unique”.
My prompt resulted in this
The visual output was surprisingly aligned with my usual image style.
What I found interesting was that ChatGPT 5.5 Thinking seemed to use reasoning together with context from past sessions / saved memories / files that included my previous prompting and image style.
So if you ever feel low on creative inspiration, it can be useful to test with simple keywords first and see how ChatGPT interprets and reasons through the prompt before generating the final image prompt/tool call.
A bit risky to shake hands with him
your image reminded me of a problem on the older models, so i thought i’d give it a go lol
I’m not really stuck on 1800’s photography, and I would love to invite other potential ideas for June…
The models are always ready to answer that question -too…
I found the offerings for ‘dev forum friendly’, slightly too well placed?
Most of the ideas are great with a little human polishing…
They’re all pretty good seeds for humans to get ideas from…
Maybe it’s just Ai slop to weary eyes…
Oh! What problem, what am I missing?
blasts from the past! haha…
i’m okay with anything for June topic. there’s not too many of us.
i like time periods… 1800s or older…
https://community.openai.com/t/monthly-or-bi-weekly-dalle-challenges/912633/3
