# ChatGPT still doesn't know how to add words to an image, even after 2 years **URL:** \ **Category:** Bugs\ **Tags:** dalle3\ **Created:** [January 2, 2025, 6:09pm UTC](https://community.openai.com/t/chatgpt-still-doesnt-know-how-to-add-words-to-an-image-even-after-2-years/1078438 "2025-01-02T18:09:20Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![oanaaaa08](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/oanaaaa08/32/758649_2.png) [@oanaaaa08](https://community.openai.com/u/oanaaaa08)\ **Post date:** [January 2, 2025, 6:09pm UTC](https://community.openai.com/t/chatgpt-still-doesnt-know-how-to-add-words-to-an-image-even-after-2-years/1078438/1 "2025-01-02T18:09:20Z")
Even now, after 2 years, DALL-E still can’t do a simple thing. **“Cel mai greu lucru este sa-ti gasesti talentul. Al doilea, sa te tii de el.”** Prompt: Create an image that contains exactly these words, in Romanian. ![image](https://us1.discourse-cdn.com/openai1/original/4X/d/5/3/d530047ca6321479f2278b72929631b5983ba150.jpeg) As you can see, it didn’t add the Romanian words correctly. If they were in English, it would have written them correctly. Basically, it has to generate the image, then remove the words somewhere there, with a certain font. Why can’t ChatGPT do that? ---
**Author:** ![polepole](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/polepole/32/740575_2.png) [@polepole](https://community.openai.com/u/polepole)\ **Post date:** [January 2, 2025, 7:04pm UTC](https://community.openai.com/t/chatgpt-still-doesnt-know-how-to-add-words-to-an-image-even-after-2-years/1078438/2 "2025-01-02T19:04:13Z")
Short answer from OpenAI Help page: > DALL·E is not currently designed to produce text. And this is long answer: [How can I generate text in my image?](https://help.openai.com/en/articles/6781228-how-can-i-generate-text-in-my-image?utm_source=chatgpt.com) ADDITION: Prompt: > A man enjoying a strawberry with the text above him: “How many ‘R’s are in ‘strawberry’?” And use lowercase letters. ![polepole-strawberry](https://us1.discourse-cdn.com/openai1/original/4X/b/1/6/b1650c0a0fe4c96312da5e476942af0ccaec230b.jpeg) ---
**Author:** ![oanaaaa08](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/oanaaaa08/32/758649_2.png) [@oanaaaa08](https://community.openai.com/u/oanaaaa08)\ **Post date:** [January 4, 2025, 8:11pm UTC](https://community.openai.com/t/chatgpt-still-doesnt-know-how-to-add-words-to-an-image-even-after-2-years/1078438/3 "2025-01-04T20:11:06Z")
yes, ok, but we need to evolve a little. Do we stay at the same stage? ---
**Author:** ![polepole](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/polepole/32/740575_2.png) [@polepole](https://community.openai.com/u/polepole)\ **Post date:** [January 4, 2025, 8:32pm UTC](https://community.openai.com/t/chatgpt-still-doesnt-know-how-to-add-words-to-an-image-even-after-2-years/1078438/4 "2025-01-04T20:32:24Z")
I think OpenAI is aware of this issue and plans to address it in future updates. But, in the competitive world of AI, companies typically avoid announcing exact release dates. **Loose lips sink ships** , sharing too much information can give competitors an advantage. Instead, they prefer to move quietly and surprise users when the time is right, because **actions speak louder than words** and delivering results matters more than making promises. **The race is not always to the swift** , I mean a success doesn’t always go to the fastest, but to those who plan carefully. Also you can visit here: > [@How to prompt to get number of object accurate](https://community.openai.com/t/how-to-prompt-to-get-number-of-object-accurate/1079735/2): > > Hi @goldenaxes.net wave Welcome people_hugging to the community! Short answer: No solution for now with DALL-E for your need. Long answer: DALL-E 3 still struggles with spatial awareness and counting objects accurately. For example, if you ask for “2 apples on the table and 3 on the floor,” the model might show a different number. This is because it has a hard time understanding exact numbers and object placement in images. The system also struggles with text rendering (like writing…