The perfect photo is in the can, but there is a trash can in the background, a stranger is walking through the frame, or a utility box ruins the look of the image. In the past, something like this meant hours of manual work with the clone stamp and the repair brush, and the retouching was often still noticeable. This tutorial shows you how to solve even difficult cases in a manageable amount of time with Affinity Photo and the AI DALL-E 2. You will learn where the strengths of the built-in fill function lie, when it is better to turn to an AI, and how to combine both tools.
The video is only loaded from YouTube when you play it – data (including your IP address) is then transmitted to YouTube/Google. More in the privacy policy
Key Takeaways
- The inpainting function in Affinity Photo delivers very good results for simple subjects against uniform backgrounds, such as grass, sand, or water.
- For complex subjects such as animals with a head or eyes, the fill function quickly reaches its limits because it doesn't know what actually belongs there.
- DALL-E 2 can fill in missing image areas appropriately, provided you cut out the area and enter a suitable keyword.
- With the restoration tool and the clone tool, you achieve a realistic transition between the original image and the AI addition.
- Save your intermediate versions so that you can compare results directly with each other later.
Step-by-Step Guide
Step 1: Roughly select the distracting object
Start with the Freehand Selection tool. You don't have to work to the millimeter, but the outline should completely enclose the object, including its shadow. In the example, a woman doing yoga on the beach is marked in this way. The more generously you select the area, the better the software can reconstruct the surroundings.
Step 2: Apply inpainting via the Fill menu
In the menu, open Edit, choose Fill, and then click Apply Inpainting. The calculation takes a moment depending on the image size. With the yoga woman, plausible structures immediately appear in the sand and water that blend well into the surroundings.
Step 3: Check the result and deselect
Press Ctrl+D to deselect. This way you see the result without distracting selection lines. Pay attention to transitions, edges, and repeated patterns. In the example, there is a small spot that looks slightly offset. You correct this with the restoration tool by carefully going over it with a small brush size.
Step 4: Recognize the limits of the fill function
Not every subject can be removed this cleanly. In the second example, you select the woman in the foreground and fill the area using inpainting as well. The result looks unnatural: the structure of the background is distorted, and there are spots where the manipulation is immediately noticeable. If you notice such errors, a pure fill is not the right tool. You can touch it up, but the effort increases significantly.
Step 5: Fine-tuning with the restoration and clone tools
Before you give up, the rework is worthwhile. The restoration tool smooths transitions when you run along the edges with a small brush. For blurry areas, such as stones in the background, the clone tool is suitable. Hold down the Alt key to set the source, and then paint the structure onto the desired spot. With this combination, the retouching looks much more realistic.
Step 6: Cut out the object and export as PNG
Before you finish the process, copy the edited image with Ctrl+C and then undo the working steps. This way you have the result ready for comparison. Select the woman again, remove her with the Delete key, and make sure the area is transparent. Then save the file via Export with Ctrl+Alt+Shift+S. Choose PNG as the format, because only this format preserves transparency. Give the file a unique name so you can find it quickly later.
Step 7: Upload the image to DALL-E 2 and define the area
Open DALL-E 2 and upload your cut-out image using the upload option. By default, the interface only generates square images. Therefore, place a square over the transparent area. Make sure that the square begins where the transparency starts. It does not have to extend exactly to the top edge of the image; that is fine. Then confirm your selection with the button for cropping.
Step 8: Enter a keyword and generate the result
Now enter a short keyword that describes what should appear in the marked area. In the example, the instructor uses the English word "trees" for Bäume. The generation takes only a few seconds, and a progress indicator shows you the status. In the result, the AI continues the grass and stones surprisingly well. If you are happy with the solution, confirm it with Accept and then save the image.
Step 9: Place the result back into Affinity Photo
Open the saved image in Affinity Photo using the Place command. A single click is enough to insert it into the document. Zoom out to assess the overall image and compare it with your copy of the original. The stones and the continuation of the grass stand out in particular in a positive way. A close comparison at higher magnification shows that the AI works significantly better here than the plain fill function.
Step 10: Remove complex subjects such as a lion
In the third example, a lion is selected. The fill function of Affinity Photo fails here because it does not recognize that an eye or the head would have to continue at this spot. The result shows exactly this uncertainty. Therefore, select the lion roughly, remove it with the Delete key, and save the image again via Export as PNG so that the transparency is preserved.
Step 11: Generate several variants and choose the best one
Upload the cut-out image to DALL-E 2 again and define the area. This time, enter the keyword "lion". The results can be surprising, and the quality in the example is impressive. You do not have to settle for the first suggestion: click through the offered variations until you like one of them. In the example, a second lion is also added in the background at the right edge of the image by adjusting the description accordingly. Then accept the result.
Step 12: Adopt the result and fine-tune it
Save the generated image and place it back into Affinity Photo. If you do not like the lion, you can remove it at any time. Use the clone tool again to smooth transitions or adjust details. This fine work often determines whether the addition looks believable.
Step 13: Cut out the cup and remove the cats
In the last example, two cats are to disappear while the cup remains the focus. Remove the cats with the selection tool and export the image as PNG. Then upload it to DALL-E 2, define the area, and enter the keyword "cup" for Tasse. The generation here also takes only a few seconds. Not every variant fits: in the example, the area at the bottom with the butterfly does not look convincing. In this case, simply generate again until a result is coherent, and then adopt the best one.
Step 14: Use AI help when you get stuck
The most important tip at the end: if you get stuck at some point because you do not know where the image information for an area should come from, use an AI such as DALL-E 2 as support. You can export the result, import it into Affinity Photo, and continue working there. This combination saves a lot of time and often delivers results that would otherwise only be possible with considerable effort.
Summary – Removing distracting objects with Affinity Photo and DALL-E 2
The inpainting function of Affinity Photo is a quick way to get started with object removal, especially with simple backgrounds. As soon as structures become more complex, for example with animals or people with recognizable body parts, the AI completion with DALL-E 2 offers the decisive advantage. What matters here is a clean cutout saved as a PNG, choosing the right area, and a fitting keyword. You finish the editing with the restoration tool and the clone tool, so that the retouching remains as inconspicuous as possible.
FAQ
Can I also remove complex subjects like animals with the fill function?With complex subjects, the fill function often reaches its limits, as it cannot recognize which structures are missing. Using DALL-E 2 is recommended here.
Why does the image have to be saved as a PNG?Only the PNG format retains the transparency of the cut-out area, which DALL-E 2 needs for completion.
Why is the area in DALL-E 2 square?The interface generates square images by default, so you have to mark the area to be completed with a square.
Can I generate several variants to choose from?Yes, you can click through the variations offered and select the one that best fits your image.
What should I do if the result is still not convincing?You can smooth transitions with the restoration tool or refine details with the clone tool. Alternatively, you can simply generate again.
Why is it worth comparing with the original?A comparison shows you whether the retouching is truly inconspicuous and helps you identify the best variant.
