Question 1

How much does each run cost?

Accepted Answer

Each watermark removal run costs 4 credits (~$0.06 on a Basic plan). That is the true cost of a gpt-image-1 edit with a small margin — we pass provider cost through rather than subsidizing it. New accounts get 10 credits on signup so you can try the tool twice without paying. Subscription plans bundle credits at a lower effective rate if you plan to use it frequently.

Question 2

Is removing watermarks legal?

Accepted Answer

It depends on the source. Removing watermarks from images you own, have a license for, or have explicit permission to edit is legal and common — e.g. cleaning up your own product photos that had preview watermarks, or restoring an old family photo. Removing watermarks from images you do NOT have rights to (stock photos, copyrighted art, someone else's work) is illegal in most jurisdictions and violates our terms of service. You are responsible for ensuring you have the right to edit the images you upload.

Question 3

How well does it work on complex backgrounds?

Accepted Answer

Very well on continuous textures (sky, water, walls, grass, skin) and moderately well on structured content (patterned fabric, repeating tiles). It struggles with highly specific details that the AI cannot infer from context — for example, removing a watermark that overlaps a unique object you need to preserve. For best results, mask tightly around the watermark and give the model as much surrounding context as possible.

Question 4

What image sizes work best?

Accepted Answer

gpt-image-1 processes inputs up to 1536×1024 natively. Larger images get resized down, processed, and scaled back up — which is fine for most uses but can soften fine detail. For portfolio or print work, start from a high-res source and consider running the AI Image Upscaler afterwards if you need 4K output.

Question 5

How much of the image can I mask?

Accepted Answer

Practically, anywhere from 5% to 40% of the image gives the best results. Too little mask and the AI has nothing to work with; too much and it has to imagine detail that is not implied by the remaining content. Watermarks typically cover 10–25% of an image which is the sweet spot.

Question 6

Do I need to be precise with the brush?

Accepted Answer

Paint tightly around the watermark with a small margin (3–5 pixels). A mask that is too loose pulls in surrounding pixels you wanted to keep, and a mask that is too tight leaves visible edges. If you make a mistake, use undo or the eraser to tighten it up before running.

Question 7

Can I remove multiple watermarks in one run?

Accepted Answer

Yes. Paint over all the regions you want removed and run once — gpt-image-1 handles disjoint masked regions in a single pass. This is more efficient than running once per watermark.

Question 8

What if the result still shows a ghost of the watermark?

Accepted Answer

This usually means the mask did not fully cover the watermark. Re-upload the same image, expand the mask by a few pixels, and run again. If ghosting persists on complex backgrounds, try opening the image in the main Studio for manual touch-ups.

Question 9

Are my images private?

Accepted Answer

Yes. Uploads are stored in your private account workspace and never shared or used for training. OpenAI receives the image and mask for processing but does not retain them per their API data policy. You can delete any upload at any time.

Remove Watermarks with a Brush

Precise removal, preserved detail

Paint exactly what goes

No more clone-stamp artifacts

Keeps the rest of the image untouched

Pay only for what you use

Three clicks from watermarked to clean

Upload the image

Paint over the watermark

Click Remove Watermark

Frequently asked questions

Need more than a quick fix?