Getting depth in cmyk conversions

[rl-retouch]

Hey all

Im retouching some cosmetic product and I'm having issue getting depth of tone when converting to cmyk. I've made some proofs and the colour is spot on, its just that the variations in highlights and gloss flatten out way too much during the conversion. I realise that it may actually not be possible to reproduce this in print but can anyone share technique on how to better this?

Thanks in advance

[andrewrodney]

Re: Getting depth in cmyk conversions

Did you try various rendering intents?
Is Black Point Compensation on?
What you report is to be somewhat expected. Some Midtone contrast (Clarity in Adobe terms) might help, you might just need to do some minor curves editing.
Open the original and soft proof with the CMYK profile, edit on layers on the RGB data and see if edits prior to CMYK will help. Once you convert to CMYK, there's less you can do although having the ability to edit just the black channel (within reason) might be useful for you. Try to fix all the issues while in RGB while soft proofing in CMYK.

[klev]

Re: Getting depth in cmyk conversions

Is it just a duller white or are you actually losing detail? It sounds like you're clipping detail somewhere. I'm assuming you're doing the conversion in photoshop. When you make the conversion, your choice of profile will take care of things such as ink limits.

Try view -> proof colors -> whatever cmyk profile you're using for conversion

Then hit gamut warning. Conversion type should be relative colorimetric, because perceptual will never work as intended.

If you will lose details due to clipping going from the original rgb image to the appropriate cmyk profile, you will see clipping. You should also ensure that they aren't just clipping due to brightness. Paper whites look weird.

Lastly make sure you're using the right profile for conversion. The wrong ink limits would end up flattening out some saturated makeup colors.

This is basic stuff, so I'm a little skeptical whether it will be helpful. I don't know if you mean almost white glossy areas or just saturated colors that appear bright. It does sound like a possible tradeoff between the right color and level of detail, but I'm obviously not sure.

[andrewrodney]

Re: Getting depth in cmyk conversions

Then hit gamut warning.

Too bad that very old 'tool' is both inaccurate and rather buggy (it will show OOG colors that are not out of gamut).
If covering your image with an overlay that treats 1% and 90% OOG the same appears to be useful to you, fine but better to soft proof and just ignore this OOG overlay!

Conversion type should be relative colorimetric, because perceptual will never work as intended.

That's a ridiculous statement that has zero basis in color management facts! For one, every profile manufacturer produces a different Perceptual Rendering, it's up to their color scientist to build one as they see fit. 2nd, if you have the tools to control the profile, you'll do so in that table and if you're the least bit interested in dealing with OBA's, again, you'll be tweaking this table. The idea that Perceptual rendering is always 'wrong' is silly and unfounded and if you believe this to be true (again it's not), you should talk to the ICC, Adobe and engineers for every product that offers a Perceptual Rendering with an ICC. Good luck there...

[klev]

Re: Getting depth in cmyk conversions

Too bad that very old 'tool' is both inaccurate and rather buggy (it will show OOG colors that are not out of gamut).
If covering your image with an overlay that treats 1% and 90% OOG the same appears to be useful to you, fine but better to soft proof and just ignore this OOG overlay!

Sure it's rough and buggy. It can help you spot problem areas. I guess you could just eyedropper it to begin with, but you're looking for 0 or near 0 values of cyan, magenta and yellow, which could result in a detail loss. I mean 0 definitely clips, but sometimes less than 3% or so can still blow out depending on the press. For proofing you shouldn't see trouble unless more than a couple pixels do hit 0%.

That's a ridiculous statement that has zero basis in color management facts! For one, every profile manufacturer produces a different Perceptual Rendering, it's up to their color scientist to build one as they see fit. 2nd, if you have the tools to control the profile, you'll do so in that table and if you're the least bit interested in dealing with OBA's, again, you'll be tweaking this table. The idea that Perceptual rendering is always 'wrong' is silly and unfounded and if you believe this to be true (again it's not), you should talk to the ICC, Adobe and engineers for every product that offers a Perceptual Rendering with an ICC. Good luck there...

Saturation and absolute colorimetric aren't designed for this. Perceptual was broken in the ICC specification itself. It has never worked as intended, so I wonder if you ever tested it. The idea behind it was that in the conversion
rgb->LAB->cmyk there may be colors that fall out of bounds. Relative colorimetric would clip them. Perceptual should compress the overall range to fit in the destination gamut. That means it only clips in the destination gamut if it clips in the source gamut, but you may lose some contrast. It doesn't work that way in its actual implementations. Open an image right now if you don't believe me.

[andrewrodney]

Re: Getting depth in cmyk conversions

Sure it's rough and buggy. It can help you spot problem areas.

Like what? And you're going to do what with this OOG area? Play with the sponge tool a bit? Or just let the profile handle OOG data as it can do faster and better in most cases:

http://digitaldog.net/files/LR4_softproof2.mov

Saturation and absolute colorimetric aren't designed for this.

Saturation is simply a RI that favors saturation over hue or brightness and it's perfectly acceptable to use IF you prefer it's rendering as viewed in the soft proof over Perceptual or RelCol. It's fair game. Abs and RelCol are IDENTICAL, it's the same damn table. The only difference is how paper white maps; Abs is intended for proofing as it now deals with media white differences.
ICC profiles and RI's know NOTHING about color in context, only humans viewing millions of pixels can do this and make judgements from there. An ICC profile only knows about a single pixel value with no context of the others.

Perceptual was broken in the ICC specification itself.

SO you say, how about some proof? It doesn't appear the color management world including geeks and engineers agree with you or this wouldn't have happened 20+ years ago. Are you suggesting that two or more decades of conversions using ICC Profiles set for Perpetual produce perhaps millons of bad images simply due to that table? Please share whatever you're smoking to come up with this flat earth concept!

It has never worked as intended, so I wonder if you ever tested it.

I test it with every image I convert and every profile I build (and I can pretty much assure you, that's more profiles than most mortals, probably you as well). Your statement is nonsensical and unless you've tested all the profile products and their differences that can be applied to a Perceptual table, you're simply blowing smoke up our rear ends . Not even worth consideration or debate. You don't want to use other RI's and force RelCol on all your conversions, go for it. Bad, very bad advise for others. Simple: toggle Saturation, RelCol, Perceptual for each image to view a soft proof of color in context. Decide what you visually prefer. Forget OOG overlay, it's buggy and ugly and the profile will handle this for you. You can produce output specific edits if necessary but that's totally different from futzing with the OOG overlay till it's gone. That's a massive waste of time.

The idea behind it was that in the conversion
rgb->LAB->cmyk there may be colors that fall out of bounds. Relative colorimetric would clip them. Perceptual should compress the overall range to fit in the destination gamut.

That's part of the basic idea and as I told you, every profile vendor can do this as they see fit. Some do it MUCH better than others. Case in point are the two different profiles Epson installs with many or their printers. One group is from Seiko Japan and have real problems mapping blues to black; their color engine sucks. X-rite created profiles of the other set supplied provide no such issues. Very good tables and profiles.
You simply cannot speak for all profiles and their perpetual RI's with such a blatant and frankly incorrect opinion about so many differently constructed RI tables!

[rl-retouch]

Re: Getting depth in cmyk conversions

Thanks Guys, im going to test it further in the morning. But to clarify its not whites that are clipping, its the subtle reflections and shape of orangey red polished nails that im losing. Most of the image is out of gamut but i dont want to reduce the saturation too much for colour matching sake. Yes im converting in PS to iso coated v2 and printing with a gmg epson. Hopefully i'll have more luck tomorrow.

[klev]

Re: Getting depth in cmyk conversions

Thanks Guys, im going to test it further in the morning. But to clarify its not whites that are clipping, its the subtle reflections and shape of orangey red polished nails that im losing. Most of the image is out of gamut but i dont want to reduce the saturation too much for colour matching sake. Yes im converting in PS to iso coated v2 and printing with a gmg epson. Hopefully i'll have more luck tomorrow.

I'll respond to Andrew later, but that sounds like you're losing detail in saturated colors. Check for clipping. If it is unacceptable, you will probably have to settle for something with better detail and slightly lower saturation. You would want to make any changes before converting to cmyk. Otherwise you lose detail to clipping, and at that point it is gone.

[rl-retouch]

Re: Getting depth in cmyk conversions

Great, it's just disheartening when cmyk destroys so much after all the hard work

[andrewrodney]

Re: Getting depth in cmyk conversions

Thanks Guys, im going to test it further in the morning. But to clarify its not whites that are clipping, its the subtle reflections and shape of orangey red polished nails that im losing. Most of the image is out of gamut but i dont want to reduce the saturation too much for colour matching sake. Yes im converting in PS to iso coated v2 and printing with a gmg epson. Hopefully i'll have more luck tomorrow.

So you're seeing this on the proof or just on-screen? You know this is the specific profile to be using?

Be useful to post a small low rez RGB version (300x300 pixels) which I could load into ColorThink and examine the actual gamut of the image itself compared to the profile. But look, the gamut is what it is, you can't do anything about it. OOG colors will clip or render based on the profile and the RI you use, you really don't need to be overly concerned with clipping per se but rather, soft proof to the CMYK profile while still in RGB! See if in the larger gamut space there's any selective color tweaks you can apply to those areas you mention. Keep in mind that you simply can't reproduce some colors so attempt to maintain the color relationships you see visually. ICC profiles, any automatic process simply can't do this. We humans have to view pixels in context and make intelligent and artistic decisions.

NEVER show the client the RGB data! Always show them the soft proof in CMYK.

[Shoku]

Re: Getting depth in cmyk conversions

Keep in mind that you simply can't reproduce some colors so attempt to maintain the color relationships you see visually. ICC profiles, any automatic process simply can't do this. We humans have to view pixels in context and make intelligent and artistic decisions.

NEVER show the client the RGB data! Always show them the soft proof in CMYK.

I agree with this one hundred percent. We get lots of out of gamut RGB files and the best way to deal with them is to adjust while still in RGB to reduce saturation in the problem areas, then convert, and then adjust again as needed.
Doing this will help maintain detail that otherwise would be lost.

Some clients have difficulty understanding these limitations because they work only in RGB.

[andrewrodney]

Re: Getting depth in cmyk conversions

There's an old saying: God (I prefer Dog) created RGB, man created CMYK, which would you rather use?

[klev]

Re: Getting depth in cmyk conversions

Great, it's just disheartening when cmyk destroys so much after all the hard work

It doesn't necessarily do so. If it's going to be shown online anywhere, they could still use that. For print it usually looks better to err on the side of preserving detail. It's not as big of a deal as it seems. Some of the camera profiles are much wider than Adobe 1998, yet none of them are an easy fit into a workflow involving photoshop.

[klev]

Re: Getting depth in cmyk conversions

So you're seeing this on the proof or just on-screen? You know this is the specific profile to be using?

Be useful to post a small low rez RGB version (300x300 pixels) which I could load into ColorThink and examine the actual gamut of the image itself compared to the profile. But look, the gamut is what it is, you can't do anything about it. OOG colors will clip or render based on the profile and the RI you use, you really don't need to be overly concerned with clipping per se but rather, soft proof to the CMYK profile while still in RGB! See if in the larger gamut space there's any selective color tweaks you can apply to those areas you mention. Keep in mind that you simply can't reproduce some colors so attempt to maintain the color relationships you see visually. ICC profiles, any automatic process simply can't do this. We humans have to view pixels in context and make intelligent and artistic decisions.

NEVER show the client the RGB data! Always show them the soft proof in CMYK.

Okay now I'm really confused. I said to tweak it before converting and you became angry. As I mentioned you should favor detail. I think hue/saturation or channel mixer combined with a mask are appropriate for something product oriented. I'll also mention that matrix based profiles don't really have lookup tables in the same sense. They're defined by basis changes from rgb to and from some pcs (typically LAB), so rgb->lab->cmyk. Lookup tables work differently. After reading this I'm not sure why you were angry before. I referred to photoshop for most of it, but perceptual is still a really bad mode. They tried to fix it with v4 profile specifications because of that.

[andrewrodney]

Re: Getting depth in cmyk conversions

Okay now I'm really confused.

I so agree!

I said to tweak it before converting and you became angry.

I didn't get angry, I called you out for some ridiculous text about perceptual rendering of ALL ICC profiles.
It's rather clear you're not paying attention to the text! IF you were, you'd see that the first post made after the OP (by ME) stated:

Open the original and soft proof with the CMYK profile, edit on layers on the RGB data and see if edits prior to CMYK will help.

Yet you think I'm angry because you said what I already did before you arrived? Really?

As I mentioned you should favor detail.

You should when that's an important aspect of the image! And further, as I attempted to point out to you, that decision must be conducted by a human with some taste and ability to view the image in context and make such decisions, NOT dismissing a profile table.

I'll also mention that matrix based profiles don't really have lookup tables in the same sense.

So what?

They're defined by basis changes from rgb to and from some pcs (typically LAB), so rgb->lab->cmyk. Lookup tables work differently. After reading this I'm not sure why you were angry before

So people who don't agree with you, people who ask you to prove your points (about how the ICC didn't get perceptual right from day one; nonsense), that's what you call angry? Maybe English is a 2nd language and if so, I'll cut you more slack.

I referred to photoshop for most of it, but perceptual is still a really bad mode.

So we should take you as a perceptual table bigot? Because you've provided zero proof or evidence of what you keep saying, despite the facts of the color science I've provided that easily dismisses this bigotry.

They tried to fix it with v4 profile specifications because of that.

Who is 'they'? Not Adobe, they have very little to do with how the ICC org as a group spec's V4 functionality, there's absolutely nothing to gain from V4 profiles as there isn't a single product I'm aware of that even supports the PRMG (the 'big deal' about V4 anyway).

Your concept that everyone should never use a Perceptual rendering has no merit. I'm not angry, I'm disappointed that you can't back up any of your claims and don't seem to be following the text well. Sorry.