Smart objects greatly increase file size

[DanK]

Lately I have more included smart objects in my workflow. I noticed that even files with relatively few layers, and all adjustments other than sharpening on adjustment layers rather than duplicates, were huge, usually almost 2 GB! These are from 5D Mark IV raw files, which are about 34 MB on average.

So, this morning, I ran a simple test. I took one image and made only three adjustments in Photoshop: I increased texture with the ACR filter, dropped the midpoint with a levels adjustment, and then sharpened with a high-pass filter. This last entails creating a new pixel layer that composites the layers below. The first time, I did this with no smart objects. These are called "base" in the table below. I then redid the exact same edits, but this time converting both the base and the sharpening layer to smart objects, which I had been doing in order to preserve the option to change them.

I then saved both of the resulting files as TIF files, keeping the layers, three ways: with no compression, with LZW compression, and with zip compression. I also saved them as PSD files. The results are:

Base no comp: 646,838
Base LZW: 692,820
Base zip: 638,669
Base PSD: 489, 821

Smart objects no comp: 1,947,229
Smart objects LZW: 1,993,211
Smart objects zip: 1,939,069
Smart objects PSD: 1,444,422

I didn't do further testing to see whether one of the two smart object layers increased file size more than the other.

A few conclusions, if this small amount of data is any indication:

1. Using smart objects can hugely increase the size of TIF files, in this case roughly tripling them.
2. If you use smart objects, PSD files are smaller than TIF files, if one preserves layers in both. This is not what I had read before.
3. As others have written, LZW is inefficient with TIF files and can even increase file size, as it did in these cases.
4. In these particular cases, the benefit of zip compression was trivial.

I had already stopped using LZW files other than in Zerene, which doesn't offer zip compression. I'll stop using it.

The big takeaways for me are:

1. to think much harder about the use of smart objects. Storage is cheap, but the nuisance of filling up storage is not trivial. For me, it would mean transferring the contents of my hard drive, including my catalog, to a new external drive; duplicating that new external drive to another one that I would have to store in another location; starting a new catalog; and retrieving the old external drive whenever I want to return to photos earlier than the change.

2. I'll probably switch to PSD files as a storage default.

Any reactions?

[Manfred M]

As a rule, I use the SmartObject as a temporary approach and keep it in place (with the storage implications) until I am done. When I finish my post-processing, I rasterize the SmartObject layer before I save the archived psd (protecting most of the adjustment layers). That being, I will sometimes collapse some of these too.

When I get to the point where the work is "finished" I am looking for a master file that I might tweak, for instance to print on different paper stocks, rather than to do a significant reworking of the image.

[xpatUSA]

<>
So, this morning, I ran a simple test. I took one image and made only three adjustments in Photoshop: I increased texture with the ACR filter, dropped the midpoint with a levels adjustment, and then sharpened with a high-pass filter. This last entails creating a new pixel layer that composites the layers below. The first time, I did this with no smart objects. These are called "base" in the table below. I then redid the exact same edits, but this time converting both the base and the sharpening layer to smart objects, which I had been doing in order to preserve the option to change them.

I then saved both of the resulting files as TIF files, keeping the layers, three ways: with no compression, with LZW compression, and with zip compression. I also saved them as PSD files. The results are:

Base no comp: 646,838
Base LZW: 692,820
Base zip: 638,669
Base PSD: 489, 821

Smart objects no comp: 1,947,229
Smart objects LZW: 1,993,211
Smart objects zip: 1,939,069
Smart objects PSD: 1,444,422

I didn't do further testing to see whether one of the two smart object layers increased file size more than the other.

A few conclusions, if this small amount of data is any indication:

1. Using smart objects can hugely increase the size of TIF files, in this case roughly tripling them.
2. If you use smart objects, PSD files are smaller than TIF files, if one preserves layers in both. This is not what I had read before.
3. As others have written, LZW is inefficient with TIF files and can even increase file size, as it did in these cases.
4. In these particular cases, the benefit of zip compression was trivial.

Any reactions?

It's always impressive when an observation is backed by actual examples.

My general comment as a non-PS user is that I'm quite astounded by the enormous difference in file size between output-referred (intermediate processed and stored) and a final output. For example, your "almost 2GB" that could (for example) be output as a moderate-quality 1600x1080px 500Kb JPEG for posting here! An "efficiency" of say 0.025 per cent!

I guess it all depends on the workflow and purpose, eh?

Er, what is a "Smart Object"?

[DanK]

Ted,

A smart object is basically a container that includes a regular pixel layer and xml instructions for parametric edits. Outside of the container, that is, in the regular photoshop interface, it behaves just like any other layer. However, you can re-open the container and change the edits at any time. So, for example, in one file that got me suspicious about this, I wanted local contrast, but I decided that I wanted to create it using the LR/ACR texture slider because its effects are subtle. If I had simply opened an ACR filter on the base layer, added texture, and accepted it, the resulting layer would have the pixels altered, and there would be no way to reverse the edit. However, by using a smart object, I could reopen the filter and change it at any point, even after a bunch of other photoshop edits. (However, if there are subsequent layers, the effects of reopening the filter are only visible after you accept the changes and close the container again.) It's very handy.

Dan

[xpatUSA]

Ted,

A smart object is basically a container that includes a regular pixel layer and xml instructions for parametric edits. Outside of the container, that is, in the regular photoshop interface, it behaves just like any other layer. However, you can re-open the container and change the edits at any time. So, for example, in one file that got me suspicious about this, I wanted local contrast, but I decided that I wanted to create it using the LR/ACR texture slider because its effects are subtle. If I had simply opened an ACR filter on the base layer, added texture, and accepted it, the resulting layer would have the pixels altered, and there would be no way to reverse the edit. However, by using a smart object, I could reopen the filter and change it at any point, even after a bunch of other photoshop edits. (However, if there are subsequent layers, the effects of reopening the filter are only visible after you accept the changes and close the container again.) It's very handy.

Thanks Dan, now it is clear ...