Outpaint vs Upscale vs Content-Aware Fill: Pick the Right Operation
A decision matrix for image editing operations — when outpaint beats resampling, when upscale beats outpaint, when Photoshop content-aware fill is still the right call. With test images and side-by-side outputs.
We built the tool we wished we had ourselves. A three-step agent that takes one photo, asks for the target shape, and adds the missing edges. But after shipping it, we noticed people reaching for outpaint when they actually needed something else, and reaching for upscale when outpaint was the right move. So here is the cheat sheet we wish someone had handed us.
Four operations. Each one changes a different thing about an image. Pick the wrong one and you waste credits, or worse, you ship a stretched, blurry, or oddly cropped result.
The Decision Matrix
| Operation | Use when | Output | What’s preserved | What changes |
|---|---|---|---|---|
| Outpaint | The subject is fine, the canvas shape is wrong | Wider or taller canvas, same subject untouched | Original pixels, exactly | Canvas size, plus new pixels at the edges |
| Upscale | The canvas shape is fine, the resolution is too low | Same shape, higher pixel count | Composition, framing | Pixel count, sharpness, fine detail |
| Content-aware fill | Something inside the frame needs to disappear | Same canvas, same resolution, cleaner interior | Canvas, surrounding context | Pixels in a chosen region |
| Crop | You have too much frame, not too little | Smaller canvas | Resolution within the kept region | Canvas shape, source content lost |
Read it as four separate verbs, not four flavors of the same thing. Outpaint adds. Upscale multiplies. Fill replaces. Crop subtracts.
Outpaint: Add Pixels Around the Edges
Outpaint changes the canvas shape and keeps the source pixels exactly. You feed it a 4:5 phone shot and ask for 16:9, and it paints new ground, sky, and walls outward from the original frame. The subject in the middle is bit-for-bit identical to what you uploaded.
Use it when the photo is good but the shape is wrong. The canonical case is a vertical phone photo that needs to become a 16:9 listing hero, a 1.91:1 social ad, or a 21:9 ultrawide banner. You did not frame for the placement when you took the shot, and now you need wider context, not zoomed-in detail.
Outpaint is also the right call for product photography that came back too tight, for stock-replacement work where the subject is locked but the negative space is wrong, and for any “this Instagram crop chopped off my friend’s head” moment. Browse aspect ratio targets to see the common ones we hit.
What outpaint does not do: it does not add detail to your subject, and it does not increase the resolution of the original pixels. If your input is 800 pixels wide, the original middle of the output is still that same 800 pixels of source data, with new pixels added beside it.
Upscale: Multiply Resolution
Upscale leaves the canvas shape alone and increases the pixel count. A 1024 by 1024 web hero becomes 4096 by 4096 for print, with sharper edges and reconstructed fine detail. The framing, composition, and content stay the same. Only the density changes.
Reach for upscale when you have the right photo at the wrong size. Web hero needs to print at 24 by 36 inches. Old family scan at 600 pixels needs to become a wall canvas. Game asset at 512 needs to ship at 2048. Phone screenshot for a slide deck needs to survive a projector.
If the output looks too small, too soft, or pixelated when you zoom in, upscale. If the output looks the right size but the wrong shape, that is outpaint, not upscale. People confuse these because both produce “a bigger file.” The difference is whether the bigger file has more canvas or more pixels-per-inch.
A common mistake: upscaling a 1:1 portrait and expecting a 16:9 landscape to fall out. It will not. You will get a sharper, larger 1:1 portrait. To change shape, outpaint first, then upscale if needed.
Content-Aware Fill: Replace Pixels Inside the Frame
Content-aware fill removes objects or patches small holes in the interior. The canvas does not change. The resolution does not change. A region inside the frame gets repainted to match what surrounds it. Photoshop has shipped this for over a decade and it is still the right tool for the job in many cases.
Use it when something inside the frame needs to leave. Photobomber behind the bride. Power line cutting through your skyline. Watermark in the corner of a stock photo you have a license for. Discarded coffee cup in the corner of a real estate listing. Lens dust spot in the sky. A small region, surrounded by similar context, that you want to disappear cleanly.
Why we did not build this into our tool: Photoshop, Affinity Photo, and the open source GIMP plugins handle interior fill well, with manual brush control over the exact region. Our agent is for the case where the canvas itself is the problem, not for the case where one specific bottle needs to disappear from a kitchen counter. Use the right tool for the right verb.
If the object you want to remove is large, complex, or sits against a busy or unique background, content-aware fill will struggle. At that point you are looking at full background replacement or compositing, which is a different tool again.
Crop: Subtract Canvas
Crop is the operation everyone forgets to consider, because it is so obvious. It removes pixels. The canvas shrinks. Source content outside the kept rectangle is gone. Resolution within the kept region is preserved. There is no AI involved. There is no generation. It is a rectangle and a delete key.
Use crop when you have too much image, not too little. A bust photo for an ID card needs the shoulders and background gone, not new pixels added. A documentation screenshot needs the browser chrome trimmed. A landscape shot has a distracting tree on the left that you do not need.
The trap: people outpaint when they should have cropped, because outpaint feels active and crop feels lossy. If you have the framing you need somewhere inside the existing canvas, crop. You will lose nothing important and you will spend zero credits.
The opposite trap: people crop tighter and tighter trying to fit a 16:9 placement, and they end up cutting off their subject’s head. That is when outpaint earns its keep. See the worked examples on our use case pages for both directions.
The Rule
Here is the default order we run through in our heads, before we touch any tool.
- Is too much in the frame? Crop.
- Is the framing right but the shape is wrong? Outpaint.
- Is the framing and shape right but it is too small or too soft? Upscale.
- Is everything right except for one specific thing inside the frame that needs to go? Content-aware fill.
Most jobs need exactly one of these. Some need two in sequence, almost always crop-then-outpaint or outpaint-then-upscale, in that order. If you find yourself wanting to do all four, step back. The source image is probably wrong for the placement and you would be better off reshooting or sourcing a better starting point.
The tool to reach for follows the verb, not the platform. Our extender agent is the right call for outpaint. Photoshop is still the right call for interior fill. Topaz, Upscayl, and the upscalers built into modern editors are the right call for resolution. The crop tool in any image viewer is the right call for crop. Pick the verb first. The tool will pick itself.