AI is becoming increasingly intelligent. Love it or hate it, it’s a tool that can make photographers’ lives a bit easier by helping eliminate unwanted objects in our photos. Skylum (who created Luminar Neo) claims that the new GenErase and GenSwap tools are super powerful and have taken these features to whole new levels in Luminar Neo. Let’s take a quick look.
Luminar Neo’s Gen Erase and Gen Swap
In the latest update, you may have seen Gen Erase and GenSwap added to the lineup of editing tools. For this post, I am mostly going to discuss Gen Erase. What’s the difference? Gen Erase is a tool for removing unwanted elements using pixels within the original photo. There is no grabbing pixels from stock photos here. It’s not overly creative but can be a powerful tool in your arsenal of editing gear. Meanwhile, GenSwap relies on AI to create pixels. You can use either for a variety of things, like extending backdrops.
At the very basic level, GenErase removes unwanted objects while GenSwap adds elements to your images. In the above image there is a slight area on camera right where the backdrop didn’t quite cover, this was fixed quite well with GenErase. On the below image I also used GenErase and even tried GenSwap, close but not gold star.
How does it work?
GenAi Rules
When it comes to GenErase and GenSwap there are a few rules to follow to make your workflow smooth. These apply to both GenErase and GenSwap especially when it comes to adding descriptive text.
Mask larger than the area required
This is an important tip, but it is also a terrific time saver. Don’t be so precious with creating your mask, you can be a little sloppy and you do not need to be so precise with your selections. This is terrific as you can roughly select your object and select more than you need.
Mask all parts of the object, shadow and reflections
The next important tip is to mask ALL of your subjects, this includes any shadows it may create and reflections. Leaving these behind could cause some unexpected results.
Avoid verbs
Avoid verbs like add, erase, move etc. Apparently, these can give unexpected results. Instead, use nouns and adjectives. Big blue sky or small brown dog etc. Be detailed in what you would like to see happen in your image. Granted this applies to Gen Swap more than Gen Erase, but it is still worth noting. I for one was always adding verbs like erase etc and then have my selection doing something unexpected.
Be patient
Remember this is still relatively new technology, so a little patience is required. I didn’t always get perfect results every time, not even the first attempt. Sometimes it was close and at times nowhere near what I wanted. You can make multiple attempts and play with different descriptive texts.
My verdict?
Keep in mind it’s still early days, and I am sure improvements will be made to continually update these features. I tried a few simple things like extending a backdrop, removing light stands and such. Sadly, it took a long time and the results were not as expected.
In the above image, I asked GenSwap to make the backdrop a solid black wall…it made it white! So why did it fail? Good question. I am on an older PC, but so ramped up with RAM and good graphics it shouldn’t have been a problem. Image size? Well, it was taken with a Sony A7RV and about 65 MB, so that could be something to do with it. Also, it took about 40 – 45 mins per image to render. I tried various different techniques and suggestions offered by Skylum, but never seemed to have the success I saw in beta demonstrations. It’s possible its my system, or it just isn’t working as well as I’d hoped.
Model/Dancer: Alyson King
