What are the use cases for this prompt?

This prompt can be used for: Correct Text.

What is the difficulty level and which AI tool should I use?

This prompt is intermediate level. It works with all AI tools (ChatGPT, Claude, Gemini, etc.).

Are there variants of this prompt?

Yes, 3 variants are available: Beginner, Intermediate, Advanced.

✍️RedactionIntermediateAll AIs

DALL-E Prompt to Correct Text

DALL-E, OpenAI's image generation model, offers a little-known feature: the ability to correct text directly embedded in an image. Whether you've generated an image containing a typo, a misspelled word on a poster, or illegible text on a marketing visual, DALL-E allows you to intervene using its editing tool (inpainting). By selecting the area containing the erroneous text and providing a precise prompt, you can ask the model to regenerate only that portion with the corrected text. This approach avoids recreating the entire image and preserves the original style, colors, and composition. However, it's important to note that DALL-E remains an image generator and not a word processor: typographic rendering can vary, and long or complex texts sometimes require several attempts. To maximize your chances of success, use structured prompts, specify the desired font, and limit the amount of text to correct in a single pass. This guide provides you with the optimal prompts for achieving reliable text corrections in your DALL-E generated images.

Paste in your AI

Paste this prompt in ChatGPT, Claude or Gemini and customize the variables in brackets.

Edit the selected area of the image. Replace the existing text with the corrected version: "[CORRECTED_TEXT]". Keep the exact same font style, size, color, and background as the original text. Maintain perfect consistency with the surrounding design. The text must be sharp, clearly legible, and properly aligned. Do not alter any other element of the image outside the selected area.

Personalize this prompt with Léa

Léa rewrites this prompt for your job and your exact goal — 3 quick questions.

Why this prompt works

This prompt works because it leverages DALL-E's inpainting feature by targeting only the area to modify, which reduces the risk of altering the rest of the image. By explicitly specifying the preservation of the typographic style (font, size, color), you guide the model toward maximum visual consistency. The instruction not to modify surrounding elements acts as a negative constraint that limits visual hallucinations.

Use Cases

Correct Text

Variants

Expected Output

The original image is preserved in its entirety, except for the selected area where the erroneous text is replaced with the corrected version. The new text adopts the same visual style as the original, with natural integration into the overall image design.

Frequently Asked Questions

Can DALL-E reliably fix text within an image?

DALL-E has made significant progress in rendering text, but it remains an image generator, not a typesetting engine. For short words (1 to 3 words), the success rate is high, especially with DALL-E 3. For longer phrases or special characters, multiple attempts may be needed. It's recommended to check every letter in the result and regenerate if the text contains residual errors.

How do I select the area to be corrected in DALL-E?

In the ChatGPT interface, open the generated image and use the selection tool (brush or rectangle) to mask the area containing the text you want to correct. Choose an area slightly larger than the text to give the model enough working space. Then, enter your correction prompt in the text bar. Via the API, use the 'mask' parameter with a PNG image where the area to be modified is transparent.

What should I do if DALL-E generates text with missing or distorted letters?

This is a common issue with image generation models. Several strategies can help: shorten the length of the text to be fixed (one word at a time if necessary), specify each letter in quotes within the prompt, request a sans-serif font which is usually rendered better, and increase the size of the selected area. If the problem persists, consider using a traditional editing tool like Canva or Photoshop for the text correction, then use DALL-E to visually harmonize the result.

Improve this prompt

Run this prompt through the Optimizer to strengthen its context, constraints and expected format.

Improve this prompt with the Optimizer

Comments

LéaAI
Pour un résultat fiable, isolez le texte à corriger en peignant un masque blanc très serré sur la zone : DALL-E respecte mieux les consignes quand le contexte environnant est protégé. Si la police est rare, remplacez « Keep the exact same font style » par « Match the font style exactly, using a similar sans-serif/serif font ».