Keeping Color Space Consistent

[RMW]

This issue is one I'm not sure how to address. My preference is to use ProPhoto all the way from Raw to saving as a Tiff using PS. Hoping someone will "clue me in" so I can be sure about how to handle this. (Especially if their name is Rodney.)
Thanks All.
Richard

[digitaldog]

The answer is yes. 16-bit, ProPhoto RGB until you need something else like say sRGB for the web.

[David Eckels]

Thanks, Andrew. And then embed sRGB in jpeg for export to web?

[Doug Gray]

Be aware that when you work in a really wide colorspace like ProPhoto RGB, when you convert to sRGB for web you might see some color shifts you didn't anticipate. Yo might want to first convert to sRGB while still in 16 bits then adjust any unanticipated color changes before saving to jpeg.

[digitaldog]

Be aware that when you work in a really wide colorspace like ProPhoto RGB, when you convert to sRGB for web you might see some color shifts you didn't anticipate. Yo might want to first convert to sRGB while still in 16 bits then adjust any unanticipated color changes before saving to jpeg.

And then, hope and pray everyone else sees the sRGB as you do (the web is still the Wild West of color consistency).

[digitaldog]

Thanks, Andrew. And then embed sRGB in jpeg for export to web?

Yes: always embed, never produce RGB mystery meat no matter what might then happen.

[GWGill]

Be aware that when you work in a really wide colorspace like ProPhoto RGB, when you convert to sRGB for web you might see some color shifts you didn't anticipate. Yo might want to first convert to sRGB while still in 16 bits then adjust any unanticipated color changes before saving to jpeg.

Yes.
If you do a colorimetric type conversion, then out of gamut colors will get clipped.
If you do a perceptual type conversion, them a couple of things could happen:
  1) You might get a generalized type of gamut compression with clipping anything that isn't compressed.
  2) You might get far too much gamut compression if ProPhoto is assumed to the source gamut.
Ideally when converting from wide gamut spaces you either need to manage the gamut compression/clipping manually,
or do more automated source image gamut aware mapping.

[kers]

If you work with skin colours - potraits: i wonder if it is a good idea to work in sRGB 16 bit-  from the start, since no coulour will be outside sRGB and so you have a more refined digital image ( sRGB is much a aller colorspace compared to prophoto)

[digitaldog]

If you work with skin colours - potraits: i wonder if it is a good idea to work in sRGB 16 bit-  from the start, since no coulour will be outside sRGB and so you have a more refined digital image ( sRGB is much a aller colorspace compared to prophoto)

The answer is no.

[kers]

sorry that was typed in a train on a very small iphone...

you say no, but why?

[digitaldog]

you say no, but why?

Pointless, counterproductive, time-consuming, and provides nothing useful.
Here's a portrait (a tight crop) that falls outside sRGB in some colors. What is gained by converting to sRGB? Nothing!
sRGB is an output color space for posting on the web to those old displays that are limited to sRGB. There are tens of millions of wide-gamut devices that exceed sRGB. Why convert to sRGB all the time in such a case?

[Doug Gray]

There are about 4,000 pixels in Andrew's image that have a deltaE ranging from 1 to 14.

These are all in the woman's shawl due to texture variations that cause high levels of R. Adobe RGB's R has a higher brightness than sRGB so when converting high R values in Adobe RGB, the sRGB red clips at 255. Because they are distributed points in the texture and surrounded by pixels that are in gamut, the clipping that occurs is really only noticeable at high magnification in Photoshop. They can be individually identified by ctrl-shift-Y.

There are additional areas on the woman that appear out of gamut with ctrl-shift-Y. Particularly large areas under the chin. However, they are only out of gamut by well under 0.5 dE. Photoshop's OOG is triggered when the RGB to RGB conversion produces any pixel value at either 0 or 255 so it's usefulness is limited and tells virtually nothing about whether the effect is material. I'm pretty sure Andrew has noted this in the past.

[digitaldog]

Certainly, the shawl and other areas of the image contain OOG colors. Yet, it is a portrait. Maybe Kers makes portraits that contain 100% skin only; I have no idea.
Here's another real portrait in this case, from raw capture. Of the nine versions, seven are raw to ProPhoto, then converted to sRGB for the web (where sRGB is kind of, sort of useful). Two are raw directly to sRGB. Can anyone tell me which are the two sRGB's visually, how they determined visually they were sRGB, and how they appear 'better' than the seven that didn't clip prior to going to sRGB?
Let's recall the question: If you work with skin colours - potraits: i wonder if it is a good idea to work in sRGB 16 bit-  from the start?
Why would I funnel from the get-to into sRGB?

[Doug Gray]

Let's recall the question: If you work with skin colours - potraits: i wonder if it is a good idea to work in sRGB 16 bit-  from the start?
Why would I funnel from the get-to into sRGB?

I would work in sRGB if my monitor was not wide gamut and close to sRGB. It's not so I don't. I generally prefer Adobe (1998) RGB or ProPhoto. The latter in the relatively rate situation where I am printing and there are printable colors outside Adobe RGB. Not common but sometimes.

And, if I was only going to produce an image in sRGB for the web with no other use I'd set my monitor for sRGB and work in 16* bits.  *(Actually its 15 in PS). Avoids tweaking in the event there are significant OOG colors converting Adobe to sRGB for the web.

[Doug Gray]

Here's another real portrait in this case, from raw capture. Of the nine versions, seven are raw to ProPhoto, then converted to sRGB for the web (where sRGB is kind of, sort of useful). Two are raw directly to sRGB. Can anyone tell me which are the two sRGB's visually, how they determined visually they were sRGB, and how they appear 'better' than the seven that didn't clip prior to going to sRGB?

Clipping from matrix profiles, whether from RAW or ProPhoto RGB to sRGB should produce identical results. Problem might be converting to Adobe RGB first, editing, then converting to sRGB. If the output is going to be sRGB, nothing wrong going straight to sRGB and editing in that space. Only appropriate for sRGB/web stuff. Not printing.

[digitaldog]

There's nothing "wrong" in converting to sRGB (other than the possibility of clipping), nor any reason to do so "from the start" as per the question.