
Say which text, and where
Point at it: "the caption across the bottom", "the price sticker on the box". If there are several pieces of text, list each one so none is missed.
Upload an image, say which text to take out, and get it back with the background rebuilt behind where the letters were.
Remove text from an image with AI: take captions, labels, and overlay lettering off your own pictures and rebuild the background behind them. First run free.
Upload at least one reference image before generating.
All four accept your image and redraw what the lettering covered. Gradients and fine textures are where they differ, so test a textured background on two.

OpenAI latest image model. Sharper instruction following, cleaner in-image text, up to 4K.

Photorealistic detail with tight prompt control. Accepts up to 8 reference images.

Highest quality Google model. Supports up to 4K resolution.

Strong typography and poster layouts. Handles dense in-image text at 1K and 2K.
Lettering sits on top of something. The edit works when the model knows what was under it.

Point at it: "the caption across the bottom", "the price sticker on the box". If there are several pieces of text, list each one so none is missed.

Sky, wall, fabric, skin. "Rebuild the plain blue wall behind it" gives the model a target instead of a guess.

Add "keep every other part of the picture unchanged". It is the clause that stops nearby details from being redrawn.

Lettering on sky, wall, or a smooth gradient disappears easily. Text over faces, fine patterns, or busy scenes is harder and may need two runs.

Use this on your own photos, screenshots, and designs. Do not use it to strip someone else’s credit line or copyright mark from their work.

To change the words rather than remove them, use the text editing workflow, which redraws the lettering in the original style.
Same generator, pointed at a neighbouring job.
Change the words instead of removing them, in the same style.
The same approach, for people and objects.
The prompt pattern behind every contained edit on this site.
The general reference-image workspace for edits of any kind.
Upgrade for larger monthly credits, faster generation, and commercial-ready output.
Try GPT-IMG with signup credits
For regular creators and light production use
Generate 50–125 images / month
For high-volume image generation workflows
Generate 120–300 images / month
Buy credits once and use them whenever you need. Purchased credits never expire.
Practical details for this generator page.
Upload the image, name the text to remove and what is behind it, and add a clause that keeps everything else unchanged. Generate and zoom in on the cleaned area.
On plain or smooth backgrounds, usually. Text over faces or detailed textures is harder, and a second pass or a different model can help.
Only from images you own or have the right to edit. This tool is for your own pictures, and it should not be used to strip someone else’s copyright mark or credit.
Yes. The text editing page redraws the lettering with new words in the original typeface and angle.
New accounts start with 30 signup credits. A 1K Standard edit on GPT Image 2.5 is 15 credits, so your first two text removal runs are covered. Images made on free signup credits carry a watermark.
Start with GPT Image 2.5, the default. FLUX 2 Pro costs the same 15 credits for a 1K Standard run, so you can compare the two on one brief. Gradients and fine textures are the weak spot, so test them on two models and keep the smoother fill.
Use the prompt panel above with the same account, credits, and history workflow.