A photo can have a gap you wish were something else: a blank patch of wall, an empty sky, a hole where a sign used to be. The free AI generative fill tool on CanvaPro lets you paint a mask, describe what should appear there, and generate a fill that tries to match the surrounding light. You do not need an account, and the usual download does not pick up a watermark from the page.
The live form accepts a JPG, PNG, or WEBP up to 5 MB. The editor includes a brush, an eraser, a size slider from 10 to 100 (starting at 50), plus Clear, Undo, Show Brush Marks, and Upload New Image. The prompt box asks you to describe what to generate in the masked area. Complete the security check and click Generate. Most fills finish in seconds. A preview appears so you can judge the seam before you download.
Paint only the area you want changed
A tight mask usually looks more natural than a loose scribble across half the frame. Use a smaller brush near edges, trim with the eraser, and Undo if a stroke goes too far. Show Brush Marks when you want to see the selection clearly. Sharper originals usually seam less at the mask edge than tiny, compressed thumbnails.
How to fill a masked area, step by step
- Upload a JPG, PNG, or WEBP under 5 MB, and only use a photo you have the right to edit.
- Brush the region to change, tidy the mask, and write a short description of what should appear there.
- Complete the security check and click Generate once.
- Check the preview along the mask edge. If the new content sits naturally, download it and keep the original as your master.
Good uses, and a clear limit
Designers add a window, a texture, or a simple prop to a still they control. Creators test a background idea inside a defined patch. Everyday users fill a gap after something has been removed. In every case, treat the result as generated content. A window that was never in the wall should not be presented as part of a real architectural photo if that would mislead someone.
When the job is to swap a named object rather than invent something in a blank patch, replace the object in the photo with a short description of what should sit there.
If the fill looks smeared
Start from a sharper file, tighten the mask, and simplify the prompt. Peak traffic or a large file can add a short wait. If you are trying to swap a discrete object — a bottle, a bag, a plant — for a different object in the same spot, the object replacer is the closer match. Its prompt is written for replacing a masked object, rather than generating any content in a painted area.
Uploads travel over HTTPS. Files are processed to complete your request and are not sold.
A leftover object you only wanted gone, not replaced, is usually a job for the object remover rather than another fill.
After you download the fill
Inspect shadows, brickwork, and any straight lines at full size. If a small region is weak, refine the mask on the original photo and generate again, rather than filling an already-processed export. Save the result with a name that describes the new content, and keep the source file beside it.
Check the join at 100 percent, and if a small patch still looks wrong, paint that area and run generative fill again on just that spot.
Filling a gap versus changing the whole photo
Generative fill works best as a local change. A window in a plaster patch, a continuation of brick, or a simple prop in a defined hole is easier to blend than a request to reinvent half the room. If you need a completely new background, a dedicated background tool is usually the cleaner path. Here, keep the mask as tight as the job allows.
Describe the fill in the same lighting as the photo. “An arched wooden window looking onto green trees, daylight from the left” gives the model more to match than “add a window.” If the first result seams poorly, look at whether the mask included enough of the surrounding texture for the model to copy.
Be careful with anything that could be mistaken for documentary evidence — a view that was never there, a sign that was never on the wall, a person added into a space they did not occupy. A useful fill for a mood board can still be the wrong file for a factual record.
If you only need a photobomber or a stray bag gone, the steps to remove unwanted objects from a photo are usually the tighter edit.
A short checklist
- Upload a JPG, PNG, or WEBP under 5 MB.
- Brush a tight mask, then describe what to generate in that area.
- Click Generate and inspect the seam in the preview.
- Download the fill and keep the original photograph.
If you are filling a gap in a series of similar photos, use the same kind of mask and the same wording so the new content does not change style from frame to frame. Consistency matters as much as a single beautiful fill.
When you are ready, mask one region, write one clear request, and save the preview only after the new content matches the light. A single clean fill is easier to use than several loose experiments.