The Smart Object has to be one of the more powerful tools in the Photoshop arsenal. If it still confuses you a little — and I know it does — check out this in-depth video on the myriad ways in which it can be utilized.
It took me a while to figure out how I could incorporate Smart Objects into my workflow. But, when I had a better grasp of how to use them, they sped up my workflow like no other tool I had learned to use up to that point. Their major and most obvious feature is that they are nondestructive, meaning that you cannot make any permanent changes to them. The simplest example of this, as Unmesh Dinda of PiXimperfect points out, is when you decrease the size of a Smart Object, no pixels are lost, so when you resize the image to its original dimensions, it looks exactly the same. If you try to do this with a regular/rasterized layer, the image will end up blurry, because data has been discarded during the decreasing. There are so many ways this comes in handy, especially for compositing.
The most frequent way in which I use Smart Objects is when I need to work on the original raw file in Photoshop. Instead of having to roundtrip from Photoshop to Lightroom or as an alternative to using the Camera Raw filter, I just open my images as Smart Objects from the start. As Dinda points out, it's as simple as double-clicking the Smart Object layer, and poof, Camera Raw pops up. I find this handiest when I'm blending layers in Photoshop and I haven't quite equalized the white balance between flash and ambient exposures.
These are just two small examples that have helped to make my post-processing more efficient. Many more are contained in the video. It's a long one, so grab a coffee and get comfortable.
What I find complete nonsense about smart-objects:
People tell me "It' non destructive and you can always make changes."
Well, if you do any cloning and masking on top if it, still you have to start from scratch or at least edit it in the way you would have with a pixel layer. Other than that, what does it help me, if I already color graded my image and then go back into the first layer to make an adjustment in camera-raw?! You just have to guess how the changes will look like in your image, as you don't see in camera-raw the current state of editing, and only the raw that lays underneath everything?
Applause for non-destructivity.
Other than that, you have less transform options (e.g. perspective warp is missing) and you can use less tools and don't work on a picture directly. that is when bad compositing happens, if you cannot unify everything. I can do some changes in a few clicks on a pixel layer, whereas my other retoucher colleagues need way more time to do it without the patch tool for example.
And still you always have the possibility to simply duplicate the background layer, keep the masks for the composite elements in your channels and go from there. Another way of non-destructive editing that doesn't cost so much time and keeps the file size small. I don't see any sense in using smart-objects, as it makes files completely messy, non-intuitive and slows down the whole editing process. I think that smart-objects are HIGHLY overrated.
Do your cloning first. Then create the Smart Object
If you have opened as an object, create a new layer, apply image, do your cloning and then convert to a Smart Object.
The thumbnail of the article is scary.