P: Allow JPEGs to be embedded, to save disk space

Report · Sep 18, 2019

Problem:
Let's say my client sends me a 4Mb JPEG file which needs retouching for a project. I open it in Photoshop, add a couple of simple adjustment layers, and save the image as a psd. The resulting file is now 65Mb. This waste of disk space is totally unnecessary, since all I've added to the image is a couple of adjustment layers. And if I'm working on, say, a magazine containing hundreds of photos, the amount of wasted space really stacks up.

Solution:
Whenever you open a JPEG in Photoshop, it appears as an 'embedded JPEG layer'. This operates a lot like a Smart Layer. You can apply effects to it, but the layer itself is not regarded as editable bitmap data (unless you rasterize it). Then, when you save it, the original JPEG remains embedded in its original JPEG format, so if you haven't added any raster layers, the file size should be only slightly larger than the original JPEG.

Report · Sep 18, 2019

Maybe you could like sidecar files: https://lightroomkillertips.com/writing-xmp-files-shouldnt/

Report · Sep 18, 2019

Regardless of if the idea is implemented, there are a couple things at play here that could be exploding your file size... and what version you are using.

1. Uncompressed .psd files vs compressed psd files. See what happens if you save as a .tif with zip compression instead?
2. Ancestor metadata bug: PS CC 2017 and I think early versions of 2018 had a bug that would add 10-15mb of corrupted metadata to any saved file. Are you using the latest version of photoshop?

Report · Sep 18, 2019

I was just going to say that this functionality sounds a lot more like basic lightroom work than photoshop...

Report · Sep 18, 2019

Thanks for the replies.

Lightroom is only relevant to very superficial retouching, but let's say you bring several JPEGs into a Photoshop document, convert them to smart objects, duplicate, flip, rotate, and warp them? If you save this as a Photoshop file, the file is huge.

But if all the original JPEGs were retained as JPEGs, the file size would only be a little larger than the original files. It's a much more efficient way of storing the images.

You suggested saving it as a zipped TIFF. Not only is this slow, but it only brings the size down from 66 megs to 50. Not very efficient, and you're needlessly converting a lossy file to raster data, then re-compressing it losslessly. If all the original images stayed as JPEGs, there's no wasted file space!

Which version do I use? CC 2015.

To explain why... I don't believe "newer" is necessarily better. New versions bring whole new bugs and troubles, and lately Adobe has been tending toward making its products uglier and darker, with light text on dark backgrounds, hideous splash screens, and horrors like the new "New Document" dialog. It really puts me off.

There's also the ever-increasing sense of "bloat". 2015 opens really fast. But I have actually tried CC 2018 and it took way longer to open. 2015 just works great for me and does all I need it to do. I will update if a feature gets added that I think may be useful. But that hasn't happened for a long time now.

The 10 Photoshop improvements I'm still waiting for are...

Smart Masks
Proper custom shortcuts (like with all Adobe's other applications)
Ability to launch the program quickly without it spinning up every hard drive on the machine
Small tool to show the current foreground/background color (ideally on the control panel)
Way better scrolling while dragging (scrolling should speed up the nearer you go to the edge)
Fractal-based upsampling tool (similar to Photozoom Pro)
Custom interpolation mode for smart objects
Much lighter interface colors, like classic Photoshop (new versions are so gloomy!)
Custom splash screens
Fix the bug which shows 16-bit documents in only 8-bit when zoomed out beyond 66.67%.
(This.)

Report · Sep 18, 2019

It will not be me who denies that sometimes Adobe makes strange strokes ... but I would say that some things that you comment are already implemented.

2. Change keyboard shortcuts? That can be done.
3. Install Photoshop on an SSD disk, the performance is greatly improved. The files in use also in the SSD and then to mechanical disks is RAID.
8. Photoshop allows you to change between four tones one of them a very light gray

In options you can also remove the home screen and recover the classic menu to create new documents.

Report · Sep 18, 2019

Jim, don't you think that if files could be saved this way, it would have long ago been implemented? There's no such thing as saving layered files out as the original jpgs. Once you've ventured outside what the jpg spec is capable of handling, file sizes are going to explode. Nothing can be done about that until Adobe, or someone, develops a completely new way of compressing and saving multilayered images.

And you start off this thread by saying:
- since all I've added to the image is a couple of adjustment layers.
To which people reply that you could solve this by using Lightroom, which can handle saving adjustments via xmp.

And your reply:
- but let's say you bring several JPEGs into a Photoshop document, convert them to smart objects, duplicate, flip, rotate, and warp them?

You're going way off the original script there.

What's the actual complaint/request? You're only interested in saving HD space? If you're doing this professionally, then it's a requirement of the job to have an abundance of available disk space for whatever the work demands. I just completed a job which required 200GB of retouched psd! HD space these days is measured in 1000's of GBs, so i don't think saving a little bit here and there is worthy of a request.
I'd be much more interested in them coming up with a format which allows for saving only the changes to a given file... so that when i have a 10GB psb, but all i've done is added an adjustment layer, only that is needing to be saved, instead of the entire 10GB.

Report · Sep 18, 2019

Champion, It seems to me that you have really hit the nail on the head. Because jpegs are encoded with a discrete cosine transformation they are not internally compatible with the way that data is stored in a psd file. Unless the adjustments are saved the way that they are in Lightroom as .xmp files or the equivalent it looks to me like programming transformation nightmare to implement a way to accomplish what Jim A is requesting. Like you say if it was easy it would have already been done.

Report · Sep 19, 2019

Problem:
Let's say my client sends me a 4Mb JPEG file which needs retouching for a project. I open it in Photoshop, add a couple of simple adjustment layers, and save the image as a psd. The resulting file is now 65Mb. This waste of disk space is totally unnecessary, since all I've added to the image is a couple of adjustment layers. And if I'm working on, say, a magazine containing hundreds of photos, the amount of wasted space really stacks up.

This is never a problem for me. If I only need simple adjustment layers, I can do much more parametric editing in Lightroom or ACR as already explained.
(Note: I am using Elements and nearly always use the 'Open in ACR' command for jpegs as well as for raws. Less than 5% of my edits need going to the pixel editor.)

I edit the jpegs in ACR (much faster to edit by batches), select all my edited batch and click 'Done'. All the edits are saved in the metadata header of the jpegs. The thumbnails in the organizer are updated, the catalog also. No creation of new "-edited" resulting files. No new jpeg compression from the orignal, which is still available. The editing 'recipe' is saved and available to apply to new files. Next time the file is opened (either from the editor or the organizer) it opens in ACR and it's easy to compare original and final versions.
Do I need a 'cooked' final version? I click 'Open' and from the editor, I can save in a version set at optimal compression for jpeg. Or, I export from the organizer (by batch) with my output-preferred file formats and parameters. I rarely need to store exported versions since I can reproduce them at will.

So, even if I decide to create a final "cooked" version in a version set, it won't weigh more than the original 4Mb.

This is my answer to the question about edits only by adjustment layers, but what about pictures which need the pixel editor? The initial edit from ACR needs further edits with layers and possibly many smart layers (think about scrapbooking or book pages). The real question now is 'do I need to keep everything, or only a final jpeg/psd/tiff version? I generally only need that for a given period. It's important to 'simplify' the smart layers to avoid gigantic psd/tiff files. Periodically, I review my big projects and I 'flatten' the layers and save as a jpeg. In the organizer I can 'flatten' the version set which keeps only the final version and deletes the intermediate versions.

Report · Sep 19, 2019

As long as Lr has no layers like Capture One, you must use Photoshop for this. Or Capture One 12.x of course, which is btw. much faster than Lr (Lr: speed a chemical rockets 21st century, C1: warp drive) and a light years ahead developing engine, specially if it come to de-noise and double specially on Fuji raw files. Have also a look at the "Refine Mask" feature of C1, it is freaking awesome good. But C1 has unfortunately it's own super ugly misconception, you can not export the tweaking you made to am xmp sidecar file like in Lr or Bridge+CameraRaw. Somehow it's a bout time to start a partner Youtube channel to "Camera Conspiracies" named "Image Editor Conspiracies" 😉

Report · Sep 19, 2019

PS: masks in Lr/CR do not offer all possibilities, I personally missing mostly HSL in Lr/CR masks. In C1 I have each layer full access to everything.

Report · Sep 19, 2019

Hey JIM A.
"Let's say my client sends me a 4Mb JPEG file which needs retouching for a project. I open it in Photoshop, add a couple of simple adjustment layers, and save the image as a psd. The resulting file is now 65Mb."

JIM,
I cannot see how a JPEG would be 4 megs. I can see how it would grow to a larger file, but a JPEG file is NEVER that large. It's usually around 4Kb not megs.

So really - there's no reason to embed JPEG's because they don't take up a lot of file size. I don't see a problem with JPEG file sizes. NO JIM, I don't agree with you on this one. No embedding needed. ta ta gotta go. I am a busy guy!

Report · Sep 19, 2019

C'mon y'all, let's be helpful here. If the OP had specified multi-layer smart object compositing from the start, it would have been helpful, and now it sounds more like a different ask... let's see if I can abstract this a bit more...

Basic request sounds like... remove rasterized layer data from smart-object layers on save.

This would have two prominent effects, IMO...
Pro: Smaller file size and faster saves, depending on how many duplicate instances of smart objects you have
Con: Longer delays on file-open because each smart object instance layer has to rebuild the rasterized layer

Which could have a lot of merit and might (I stress might) be relatively simple to implement.

Report · Sep 19, 2019

Why aren't you saving back out as JPEG? I'm not seeing how there is a problem here.

Report · Sep 19, 2019

OK, I had time to do an A:B:C test and here's some solid numbers.

I saved out three versions of the same base file:

4000x4000px PSD file saved from CC2019
Added one 1334x750px (240kb) JPEG as a smart object
4x copies of the smart object with varying scale and rotation transformations applied.

A) Saved as-is.
B) Saved with a solid color overlay fx of white applied to each layer (to see if that changed the rasterized save data to be smaller)
C) Saved over the jpeg inside the smart object with solid white (simulating a near total clearing out of image data from the main document's layers)

A: 125mb
B: 97mb
C: 5.51mb

So there is a clear and substantial difference in file size when the rasterized layer data is "removed".

I can't speak to how speedy the file would load if you have to re-raster all the layers but it might be worth the trade-off for people who have to do all their work from a contractor laptop with a single ssd.

Report · Sep 19, 2019

nice test. Thanks.

Report · Sep 19, 2019

Jim, Max,

125 Mb for a JPEG? What's the length of the photo, 9 blocks long?

Even if you saved that embedded, it would take as much time to save it so it does not make any sense. Oh and BTW, being a Windows engineer, I need to tell you that Windows does not dual-task, so your file save will be the time it save as you disrupt other apps on the system.

And what are you looking to save in time? 1/10th of a second on the saved file? Now multiply that by the number of the users who would be interested, and tell me if it's worth Adobe's time to make that happen? Not if it doesn't work in the first place it won't happen. It's not worth it.

Do you have a bus to catch? Time is money? What's the difference will saving 9/10th of a second help you? What other tasks do you need to get to that's more important?

If Adobe made this change, developers will keep asking themselves "what are we doing this for again??"

If you can make sense out of it, maybe they will perk their interest but I con't see any interest with a 9/10th of a second difference. And then, developers need to put 2 months into the programming and test engineers, then there's the expense for it all. It's not justified for the expense or time.

Steve Lehman, mcse

Report · Sep 19, 2019

sounds to me like what they're asking is for being able to save a multi layered psd with SOs and adjustment layers as a jpg? The only way i could imagine such a thing is that basically the psd would be a container for a layered file which contains only the adjustments and then a blank linked layer that references the jpg which is contained in the new container format.
This is really very specific use kind of thinking and is a real waste of our time to even think about.
A better approach for Adobe would be developing a resolution independent platform which we had back in the 90s with Live Picture.
But spending resources trying to save a not-jpg as a jpg? Wasn't there a season of Silicon Valley which flirted with this idea?

Report · Sep 19, 2019

Eartho I can't believe I am replying to your suggestion but it's good. But container formats are used for video. Not many containers are being used for JPEG files.

There is one which Adobe uses for illustrator with .flt (FLT)which is a filter that acts as a container format for compressed files. Definition: stands for FiLTer; a file extension used in Adobe Illustrator, PageMaker; a plugin container; not directly viewable.

This may not be used for JPEG as a container in place of utilizing embedded file formats. It may be useless info but worth a mention. It's the only active container left in my list of 200 or so file formats. All others are container files for vids.

Report · Sep 19, 2019

Steve. Read before trash-posting. I laid out the file specs

I saved out three versions of the same base file:
4000x4000px PSD file saved from CC2019
Added one 1334x750px (240kb) JPEG as a smart object
4x copies of the smart object with varying scale and rotation transformations applied.

When I mention save times, it's as a side bonus. That's totally beside the point of the OP. But since you are a Windows engineer, you have to admit that it will necessarily become faster if you drop the rasterized data from those layers. The number of bytes you are processing is much lower. I have files that can take a minute or more to save. No that won't be saving me a hours of time, but I wouldn't snappier saves.

I also noted the load time as a theoretical disadvantage because I have a lot of experience with how long it can take to rebuild the layer previews when I update and save a smart-object.

Report · Sep 19, 2019

The OP suggests embedding the JPEG as a JPEG (which is already literally how embedded smart obejcts work) and never says the main doc should be saved as JPEG with layers... it is a bit unclear... but I read it as how he introduced the example:

I open it in Photoshop, add a couple of simple adjustment layers, and save the image as a psd.

and

let's say you bring several JPEGs into a Photoshop document, convert them to smart objects, duplicate, flip, rotate, and warp them? If you save this as a Photoshop file, the file is huge.
But if all the original JPEGs were retained as JPEGs, the file size would only be a little larger than the original files. It's a much more efficient way of storing the images.

which is all how photoshop already works... the large file size comes from the rasterized pixel data in each layer that references the already very small actual embedded jpeg. As I proved in my test below...

So if you emptied out the rasterized pixel data on every SO layer before saving, you'd get the base file overhead + the tiny jpeg being compressed and saved.

Report · Sep 19, 2019

I would also point out that Adobe has a vested interest in making file sizes more efficient with the big push to cloud documents and CC Libraries.

When you start thinking of file size in relation to having to save up and down from the cloud every time, the difference between 5mb and 125mb looks a lot more significant.

Report · Sep 19, 2019

Max, "Steve. Read before trash-posting. I laid out the file specs"
Gee thanks for that. I am only trying to tell you something better.

They will never change out of rastor. And, I am not trash talking. I think others carried my thinking as well about this. I am always direct to the point as all engineers at Microsoft are this way. Apparently I am new to you, so sorry if my direct response hit you wrong, but I think most of us have been trying to tell you the same thing.

Your file may have many layers or it's big but it won't save any quicker without a container file and we haven't used those since 2005. I have a list of 200 file formats including containers and only 7 of the containers are for video only, not for your regular files, and embedded files could be as bulky and slow your system. An embedded operation can make your software more bulky and with 200 or more of those files. Your software will begin to load slower. I am not sure if you know this, but you may not be happier with its result even if there was a change in the software, as it may disappoint you greatly.

Report · Sep 20, 2019

I find none of those to be useful and I'd rather that development time wasn't wasted there.

Report · Sep 20, 2019

"They will never change out of rastor", I think you are confused about what I am suggesting... I'm not saying Adobe should change to some non-raster format, or even change anything about how they embed smart objects.

What I am saying is that when a file is saved, completely omit the pixel data from smart object layer instances. Don't save the layer pixels at all. When you re-open the file rebuild that raster data again.

"Your file may have many layers or it's big but it won't save any quicker without a container file and we haven't used those since 2005.", Yes, it will save faster, assuming the process above. Fewer pixels to compress and write when saving. Especially if you have a lot of very large, duplicated smart object layers referencing the same object, like you might find in a print ad with heavy text effects.

I'm not suggesting we change to containers or new file formats or anything except zeroing out the pixel data on smart object layers when a file is saved and triggering a refresh when loaded. With all the pros and cons that might come with that.

Report · Sep 20, 2019

3. Install Photoshop on an SSD disk, the performance is greatly improved. The files in use also in the SSD and then to mechanical disks is RAID.

Of course I use an SSD. Who doesn't in 2019!? But it's no excuse for sloppy coding. The difference in speed between CC2015 and CC2018 was really significant. Plus CC2018 had an annoying bug on my computer that made dialog boxes invisible in full-screen mode. I was very happy to revert back to CC 2015!

8. Photoshop allows you to change between four tones one of them a very light gray

The lightest option is still darker than it used to be. Plus I'm nervous that Adobe will remove the lighter options completely, like they did with Premiere. Adobe products are tending more and more towards darkness.

2. Change keyboard shortcuts? That can be done.

I mean full customization, like in other Adobe products. Currently, shortcuts are severely limited. I list those limitations here:
https://feedback.photoshop.com/photoshop_family/topics/please-can-you-allow-fully-customizable-keybo...