Lightroom’s new Generative AI Removal tools are exciting but can be unpredictable, so it’s important to know how to use them effectively in different situations.
This informative video by Anthony Morganti shows you how to best use Lightroom’s Generative AI Removal tool. Morganti starts by showing you how to remove sensor blemishes and water droplets from your images. To do this efficiently, he suggests turning off Generative AI and Object Recognition, which will allow the tool to work as before, using surrounding pixels to replace the unwanted blemishes. This method is quick and effective for minor corrections.
For more complex edits, like removing a person from a photo, the approach changes. Morganti starts by showing how the tool would struggle without generative AI and object recognition; results often include mismatched textures and overlapping elements. He then turns on generative AI and object recognition to demonstrate how these features help the tool better understand and fill in the gaps: the AI references millions of images to create more realistic replacement images, allowing you to choose from multiple variations.
For more complex scenarios, like removing a person from a detailed background, Morganti recommends adjusting the masking technique. If the mask resembles the shape of a person, the AI may replace the person with another. Using a rectangular mask instead encourages the AI to fill it in with more appropriate elements from the background. This method often produces better results and avoids the weird replacements that can occur with object-shaped masks.
There are situations where you might need to test both methods. Morganti gives the example of a photo of two people in front of a tram, where he tries both ways: recognizing the object and not recognizing it, and explains how the results change and why experimentation is important. Sometimes the changes in the AI will look realistic, and sometimes they won’t, so it’s important to be flexible in your approach. Watch the video above to hear Morganti explain in more detail.
