Why text can go wrong
AI image generation sometimes produces text on posters, signs, book spines, or other props that is misspelled or garbled. This is a known limitation of image generation models.
Best approach: remove the text entirely
The simplest and most reliable fix is to remove or replace the prop that contains incorrect text. This avoids the complexity of trying to edit text in place.
✅ Good prompt: Remove all the posters on the walls |
This is the cleanest approach. Simply remove the items that contain incorrect text and let the AI fill the space naturally.
If you must edit text
If you need specific text to appear, specify the exact target area in the image and the exact text you want, in quotes. Be precise about which element you’re targeting.
✅ Good prompt: for the left hand poster on the wall make the top text exactly "Murphys" and the bottom text beneath the glass say "Pubs & Bars". Keep the rest of the scene exactly the same |
This works because it identifies the exact poster (left hand poster), specifies the exact location of each text element (top text, bottom text beneath the glass), puts the desired text in quotes, and keeps the rest of the scene the same.
❌ Bad prompt: make the poster say Murphys Pubs & Bars |
This is too vague: which poster? Where exactly should the text go? How should it be formatted? The more ambiguous your instruction, the less likely the AI will get it right.
Recommendation
For the most reliable results, we recommend removing props with text rather than trying to fix them. If specific text is critical to your image, consider adding it in post-production using a design tool.