Hi all,

there's one thing i don't understand, may be a misconception: why is it necessary to have separate modes for editing the RGB data and the plates?

For example, if i have an RGB image in the composition and want to apply 'value curves', that has to be done in the RGB area, for after separation the plates are treated individually. Now only after manually pulling over the 'press projection' again i can discover that this operation
drove my plates out of gamut.

Is it correct that there's no 'live preview' of the effects that RGB manipulations have on the plates?

greetings, peter

On Fri, Jun 19, 2009 at 8:05 PM, yahvuu wrote:

there's one thing i don't understand, may be a misconception: why is it necessary to have separate modes for editing the RGB data and the plates?

For example, if i have an RGB image in the composition and want to apply 'value curves', that has to be done in the RGB area

In 21st century there are no RGB curves :) There are LAB and L*c*h* ones ;)

Alexandre

On Fri, Jun 19, 2009 at 10:37 AM, peter sikking wrote:

guys,

the second part of my lgm talk is blogged now:

I like this approach.

I have a few questions:

Will each plate have a density or opacity attribute? (some inks are more opaque than others)

Will it be possible to edit an individual plate in grayscale?

And finally, will it be possible to perform operations on the RGB portion of the image that do not take (immediate) effect on the projection? For example, if I want to go back and add a portion of my RGB artwork to a plate, I might want to clone and existing RGB layer, perform some modifications, then apply the contents of that new layer to one of the plates.

Thanks,
Chris

On Fri, Jun 19, 2009 at 6:56 PM, Alexandre Prokoudine wrote:

On Fri, Jun 19, 2009 at 8:05 PM, yahvuu wrote:

there's one thing i don't understand, may be a misconception: why is it necessary to have separate modes for editing the RGB data and the plates?

For example, if i have an RGB image in the composition and want to apply 'value curves', that has to be done in the RGB area

In 21st century there are no RGB curves :) There are LAB and L*c*h* ones ;)

Not that the difference really matter as both linear light RGB and CIE Lab are fully mutually transformable into each other. For CIE Lab or Lch it would only make proper sense to control the L, other controls can be used for tuning white point etc in those color models though.

/Øyvind K.

On Fri, Jun 19, 2009 at 12:41 PM, peter sikking wrote:

Chris Mohler wrote:

I like this approach.

I have a few questions:

Will each plate have a density or opacity attribute?  (some inks are more opaque than others)

I guess that the complexities of ink simulation start showing here.

Yes - although transparency/opacity would be enough for me to use GIMP for professional separation work. I'm reasonably sure that PS uses a variant of the blending mode Multiply for spot channels, and this works fairly well.

Will it be possible to edit an individual plate in grayscale?

well, as pippin said: the individual plates _are_ grayscale/monochrome drawables. they will be editable just like layer masks or selections.

Excellent.

And finally, will it be possible to perform operations on the RGB portion of the image that do not take (immediate) effect on the projection?  For example, if I want to go back and add a portion of my RGB artwork to a plate, I might want to clone and existing RGB layer, perform some modifications, then apply the contents of that new layer to one of the plates.

I am curious why you want to do something like that, because you are then going against the grain of the whole plan: freedom to develop the artistic concept further without (much) rework on the plates.

Imagine I'm designing a black t-shirt with say five spot colors, including white. After completing the artistic design, I enable the 'projection screen'. This theoretically would result in my five "plates". However, the white plate will need special attention.

Here's my workflow for this in PS: I would use the (badly named) 'Apply Image' command to take the contents of each color plate and combine them into the white plate using the mode 'multiply'. I would also manually "choke" the white plate - this means making the white areas a point or two smaller than the colored areas, thereby preventing the white from poking out at the edges of the colored areas. This process can get a bit tricky, especially if the original artwork is very complex. Often, create temporary layers (or plates), perform selection/drawing functions, then combine the result back into a plate in one of two ways - either making a selection on the temp layer and going to the plate and filling or erasing, or using the 'Apply Image' command to take the RGB channel of the current layer and combine it with a plate using a mode such as Multiply, Screen, or Add.

Now, I am quite interested in learning new workflows - so I am not bound to the "how" of the method above, but I hope I have explained the "why" well enough. In addition to being able to interact with each plate as a grayscale drawable, it would be useful to create temporary areas for doing work - be they layers, channels, plates, whatever - on which to create paths, selections, etc to in turn use to modify the plates manually. Icing on the cake would be a mechanism to combine/subtract plates using the available blending modes. During the process, it is fairly critical to have an ink density/opacity setting for each plate, to simulate (roughly) how the final print is going to look. EG, set the white plate at approx 90%, the colors at approx 70% - and you can see which portions of the colors are falling on the white underlay, and which portions are falling on the black shirt.

I realize that this is just one corner case, but if you visualize each plate being printed separately, in order, you may be able to recognize some of the many 'gotchas' inherent in separating the artistic artwork into something suitable to send to the press. That's why (in my opinion) it is important to have as much control as possible over each plate.

Whew ;)

If I explained any of this poorly, I am sorry and will happily try to do better.

All in all, I am very pleased with the direction that this is taking and I would certainly like to use GIMP for even more of my production work. :)

Thanks,
Chris

On Fri, 19 Jun 2009, peter sikking wrote:

guys,

the second part of my lgm talk is blogged now:

enjoy,

I do appreciate your work on this, but I have to say that I still have some concerns.

I have to go somewhere, so I haven't read evreything yet, but I'd like to make this point. Converting from CMYK to RGB and vice versa are not lossless by a long shot. This is by necessity and the fact that GEGL is non-destructive has nothing to do with it, since large porions of image data will have to be tossed when converting.

The gamut for RGB is far larger in the bright colors, and CMYK can produce effects that cannot be produced in RGB.

See, for example:

I am very concerned that your overlay concept would simply degrade the image over successive conversions.

Andrew A. Gill wrote:

On Fri, 19 Jun 2009, peter sikking wrote:

guys,

the second part of my lgm talk is blogged now:

enjoy,

I have to go somewhere, so I haven't read evreything yet, but I'd like to make this point. Converting from CMYK to RGB and vice versa are not lossless by a long shot. This is by necessity and the fact that GEGL is non-destructive has nothing to do with it

Isn't it a good idea to first read what you are commenting on?

It seems to me you completely misunderstood the whole thing. What makes you think there is any CMYK -> RGB conversion involved here?

/ Martin

hi,

peter sikking schrieb:

(peter) yahvuu wrote:

there's one thing i don't understand, may be a misconception: why is it necessary to have separate modes for editing the RGB data and the plates?

mainly because creating art on a RGB monitor, to be used on many media, is not the same _activity_ as bringing this art to _one_ particular printing press.

also, it is better when the art itself is separated from the adaptation of it for one press run.

ah, that helps. So it is deprecated to create single-ink 'light blue text' directly on the 'light blue' plate, as the text would then not be known to be part of the artwork.

Can you give an outline how the print color for that text will be specified? The RGB color isn't useful here and the text layer can't be accessed while the press projection is pulled over, IIUC. So each artwork layer will have a custom color separation setting resp. color mapping?

For example, if i have an RGB image in the composition and want to apply 'value curves', that has to be done in the RGB area, for after separation the plates are treated individually.

the question is what is the curve for? is it artistic? then RGB is your space. getting the plates right? then chain them together and do a curve.

Now only after manually pulling over the 'press projection' again i can discover that this operation drove my plates out of gamut.

there is going to be no substitute for experience with printing presses and especially the particular press one is working towards.

sure. Gamut warnings can only ease the first step, by guiding the color separation process.

Is it correct that there's no 'live preview' of the effects that RGB manipulations have on the plates?

if that is really needed by some users (see the two activities above and also "no substitute for experience") then a second view of the file (View->New view) can be run with the projection pulled down. might be hard on the processor.

Regarding the use-case of matching one color from a photo with a predetermined ink combination, i think this will be useful.

greetings, peter

Am Samstag, 20. Juni 2009 schrieb yahvuu:

[...]

Can you give an outline how the print color for that text will be specified? The RGB color isn't useful here and the text layer can't be accessed while the press projection is pulled over, IIUC. So each artwork layer will have a custom color separation setting resp. color mapping?

From Peter's blog post:

"However there will be full flexibility to map the content of any layer directly to any plate. For instance that light blue text in our example: it can be directly mapped from its text layer to the light blue plate, bypassing the composite."

I understand that you can use layers as plates.

hi,

Chris Mohler schrieb:

Imagine I'm designing a black t-shirt with say five spot colors, including white.

[..]

Whew ;)

Whew, too ;) Makes me wonder if it has to be that hard or if it points to some missing software improvements. Trying to understand the example, i hope you don't mind some uninformed questions (and also some out-of-sequence quoting).

Besides anticipating printing press idiosyncrasies ('choke'), it seems to me you're manually creating kind of a color separation. Quite naively: doesn't photoshop know you're printing on black?

Here's my workflow for this in PS: I would use the (badly named) 'Apply Image' command to take the contents of each color plate and combine them into the white plate using the mode 'multiply'.

this is to create the white underpinning, resp. the beginning thereof. 'Apply Image' is short-hand for 'blend anything with anything', but doesn't do any tricks that could not be achieved with layer stacks in combination with proper channel masking. On track?

I would
also manually "choke" the white plate - this means making the white areas a point or two smaller than the colored areas, thereby preventing the white from poking out at the edges of the colored areas. This process can get a bit tricky, especially if the original artwork is very complex.

if the artwork was fully vectorized, say a pure inkscape job, would that make things easier?

Often, create temporary layers (or plates), perform selection/drawing functions, then combine the result back into a plate in one of two ways - either making a selection on the temp layer and going to the plate and filling or erasing, or using the 'Apply Image' command to take the RGB channel of the current layer and combine it with a plate using a mode such as Multiply, Screen, or Add.

i assume the temporary layers are mostly grayscale?

the temporary layers serve as 'mixing stage' because it takes several steps to create a desired mask, or is it more to keep selections/drawings for reuse?

thanks for your patience, peter

Hello Martin!

On Saturday 20 June 2009, Martin Nordholts wrote:

It seems to me you completely misunderstood the whole thing. What makes you think there is any CMYK -> RGB conversion involved here?

/ Martin

I think Andrew referred to this part from Peter's article:

what about CMYK files?

When a received CMYK file is to be used in new creative work, we already saw that ‘it needs to be imported and converted to RGB.’

Daniel

Sorry, I just saw that Guillermo cleared up that misunderstanding already. (Though it was not detected as part of the same thread by my mail client.)

Daniel

hi,

Tobias Ellinghaus schrieb:

Am Samstag, 20. Juni 2009 schrieb yahvuu:

[...]

Can you give an outline how the print color for that text will be specified? The RGB color isn't useful here and the text layer can't be accessed while the press projection is pulled over, IIUC. So each artwork layer will have a custom color separation setting resp. color mapping?

From Peter's blog post:

"However there will be full flexibility to map the content of any layer directly to any plate. For instance that light blue text in our example: it can be directly mapped from its text layer to the light blue plate, bypassing the composite."

I understand that you can use layers as plates.

yep, that's the desired effect here. For the single-ink text example, a bit more than filling the light blue plate to 100% is required, though, as the other plates have to be set to 0% for that area.

My question is how and where such a layer->plate mapping will be specified, for example when i want my text to be (0, 0, 50%, 50% ,0) on the five plates.

greetings, peter

On Sat, Jun 20, 2009 at 11:27 AM, yahvuu wrote:

hi,

Chris Mohler schrieb:

Imagine I'm designing a black t-shirt with say five spot colors, including white.

[..]

Whew ;)

Whew, too ;) Makes me wonder if it has to be that hard or if it points to some missing software improvements. Trying to understand the example, i hope you don't mind some uninformed questions (and also some out-of-sequence quoting).

Besides anticipating printing press idiosyncrasies ('choke'), it seems to me you're manually creating kind of a color separation. Quite naively: doesn't photoshop know you're printing on black?

Yes - I end up doing a lot of it manually, and no it does not know - having a 'target' or 'base' would be a step forward.

Here's my workflow for this in PS: I would use the (badly named) 'Apply Image' command to take the contents of each color plate and combine them into the white plate using the mode 'multiply'.

this is to create the white underpinning, resp. the beginning thereof. 'Apply Image' is short-hand for 'blend anything with anything', but doesn't do any tricks that could not be achieved with layer stacks in combination with proper channel masking. On track?

Yes.

I would
also manually "choke" the white plate - this means making the white areas a point or two smaller than the colored areas, thereby preventing the white from poking out at the edges of the colored areas.  This process can get a bit tricky, especially if the original artwork is very complex.

if the artwork was fully vectorized, say a pure inkscape job, would that make things easier?

Of course, but when photographic-type artwork comes into play, it's usually easier/faster to do the whole thing in a raster editor.

Often, create temporary layers (or plates), perform selection/drawing functions, then combine the result back into a plate in one of two ways - either making a selection on the temp layer and going to the plate and filling or erasing, or using the 'Apply Image' command to take the RGB channel of the current layer and combine it with a plate using a mode such as Multiply, Screen, or Add.

i assume the temporary layers are mostly grayscale?

Usually RGB layers, or grayscale channels.

the temporary layers serve as 'mixing stage' because it takes several steps to create a desired mask, or is it more to keep selections/drawings for reuse?

A little of both. Sometimes I just need a very complex selection, but I need to do some work to create the selection. Other times I need to store a selection for later use (that's generally when I make an extra channel).

After re-reading the notes on the talk, if we have a Layer->Plate mapping, I think that will cover most situations. I would prefer a way to "mix" the plates, and to be able to add new layers that could later be applied to new or existing plates, but this could be worked around.

Chris

hi,

Chris Mohler schrieb:

On Sat, Jun 20, 2009 at 11:27 AM, yahvuu wrote:

i assume the temporary layers are mostly grayscale?

Usually RGB layers, or grayscale channels.

sorry, imprecise question. I mean, if you use a temporary RGB layer, it's content will still usually be just grayscale, effectively used as a mask. Assumed correctly?

greetings, peter

On Sat, Jun 20, 2009 at 12:07 PM, yahvuu wrote:

hi,

Chris Mohler schrieb:

On Sat, Jun 20, 2009 at 11:27 AM, yahvuu wrote:

i assume the temporary layers are mostly grayscale?

Usually RGB layers, or grayscale channels.

sorry, imprecise question. I mean, if you use a temporary RGB layer, it's content will still usually be just grayscale, effectively used as a mask. Assumed correctly?

Not necessarily - consider another scenario: mixing two spot colors to create a third color. More accurately, consider an image (RGB) that is primarily three colors, and then attempting to print that image in two colors, getting as close to the original RGB artwork as possible (the number of spot colors is a major factor in printing cost).

In that scenario, I may end up duplicating the original RGB artwork several times before getting the selection that I want - then ultimately filling or erasing that selection on the spot channel(s). The process can be very non-intuitive at times, and sometimes you just have to dive in and do some trial and error - esp. if the original RGB image is something flat that was not originally intended to be printed in such a way or is "damaged" (eg, not on layers, suffers from JPEG compression, etc.).

Chris

On Saturday 20 June 2009 09:40:22 am Daniel Hornung wrote:

Hello Martin!

On Saturday 20 June 2009, Martin Nordholts wrote:

It seems to me you completely misunderstood the whole thing. What makes you think there is any CMYK -> RGB conversion involved here?

/ Martin

I think Andrew referred to this part from Peter's article:

what about CMYK files?

When a received CMYK file is to be used in new creative work, we already saw that ‘it needs to be imported and converted to RGB.’

And I am not sure that this is the correct approach. Why would this be needed? Is this so that we can deal with GIMPs limited functionality to handle anything beyond RGB color spaces? If so then the focus should be on supporting other color spaces directly.

I also have not any anything in the thread related to how color management fits. After all how do you create the printer specific separations from an RGB (or other non-device specific) image into the correct device values without involving the color management engine?

Hal

On Sun, Jun 21, 2009 at 11:22 AM, peter sikking wrote:

Imagine I'm designing a black t-shirt with say five spot colors, including white.  After completing the artistic design, I enable the 'projection screen'.  This theoretically would result in my five "plates".  However, the white plate will need special attention.

Here's my workflow for this in PS: I would use the (badly named) 'Apply Image' command to take the contents of each color plate and combine them into the white plate using the mode 'multiply'.

this looks analogue to me to black generation in cmyk, but now inverted because we are on a black background. interesting concept, the media color (makes note). since it is going to work for black, it can be made to wok for any other color, with reversed logic also.

Yes, I think media color should be taken into account when designing this system - even though I suspect the vast majority of media will be white or very light.

 I would
also manually "choke" the white plate - this means making the white areas a point or two smaller than the colored areas, thereby preventing the white from poking out at the edges of the colored areas.

this looks like trapping to me. is there a difference? trapping set-up for each plate would be in the projection set-up.

A "choke" is a trap of negative amount. This is probably just jargon - I suspect that it should in fact be called a negative trap. Automatic trapping (and overprinting) has never lived up to my expectations - I would love to hear from anyone who has used auto-trapping software with acceptable results though.

Now, I am quite interested in learning new workflows - so I am not bound to the "how" of the method above, but I hope I have explained the "why" well enough.  In addition to being able to interact with each plate as a grayscale drawable, it would be useful to create temporary areas for doing work - be they layers, channels, plates, whatever - on which to create paths, selections, etc to in turn use to modify the plates manually.

everything of that will work on plates like working on layers today. I am sure that global concepts like paths and selection will be applicable to layers and plates without limits. a selection created on a layer and applied to a plate: sure.

OK - thanks for clarifying.

 Icing on the cake would be a mechanism to combine/subtract plates using the available blending modes.

to generate plates from channels/layers that is needed, but generating plates from plates? sounds like a creative kind of workflow to me.

I remember one specific instance: printing two blue colors - one light, one medium - on very dark blue. We originally placed the light blue color behind the medium blue color (overprint). The client changed their mind, and I needed to remove the overprint. Merging the (inverted) contents of med blue into the contents of lt blue removed the overprint in one step. I basically masked one plate with another and applied the mask.

While I doubt that function is necessary, it would certainly be very useful on occasion.

 During
the process, it is fairly critical to have an ink density/opacity setting for each plate, to simulate (roughly) how the final print is going to look.  EG, set the white plate at approx 90%, the colors at approx 70% - and you can see which portions of the colors are falling on the white underlay, and which portions are falling on the black shirt.

hmmm, tricky that one. it is natural for the plate stack to work sort-of like the layer stack. eye symbols to switch plates on/off. then there is the opacity slider of the layer stack. coverage slider for the plates? ay be does the dual purpose of previewing like you need and absolute print balancing.

Indeed - the stack of plates should function more or less like the layer stack. Yes - I envision a visibility toggle for each layer, and also an opacity slider. But here's another murky area (as if we needed more ;) - if I set a plate's opacity to 50%, does 100% black on that plate print out at 50% or 100%? I would expect 100% - but that's from past experience, and not very intuitive. Perhaps you are right that we need both a opacity and coverage control - that makes more sense to me, but I have never seen it implemented and may well prove confusing.

After re-reading the notes on the talk, if we have a Layer->Plate mapping, I think that will cover most situations.  I would prefer a way to "mix" the plates,

"mixing" channels + layers -> plates is a starting point for the development of the design of the plate set-up.

OK - thanks.

and to be able to add new layers that could later be applied to new or existing plates, but this could be worked around.

add layers where, image side or press projection side?

My guess is image-side. One possible scenario:

1. Design artwork in GIMP - RGB, 3 colors, 1 color per layer - 3 layers (or maybe 4 with a bg color)
2. Create print projection, map layers to plates 3. Done, hit print/export - OR
4. Go back to RGB, duplicate two layers, merge them, apply curves, etc - whatever needs adjustment
5. Manually apply the contents of the new layer to one or more of the plates in the projection
6. Done, print/export

I guess to summarize: in addition to the initial layer(or color?) -> plate mapping, it should be possible to re-apply contents of one or more RGB layer to the plates without re-mapping the entire projection (if that makes sense).

Things like overprints and trapping can get very complicated, esp if the colors are not solid and/or you are mixing spot colors. Often fine-tuning is required. I would love to see automatic trapping (complicated!), but not without being able to manually tune the results

I'd like to thank everyone for participating in this discussion! I like the direction that this is headed... Chris

peter sikking wrote:

Chris Mohler wrote:

I would
also manually "choke" the white plate - this means making the white areas a point or two smaller than the colored areas, thereby preventing the white from poking out at the edges of the colored areas.

this looks like trapping to me. is there a difference? trapping set-up for each plate would be in the projection set-up.

Note that "trapping" has two meanings in the printing world.

One relates to the way the inks stick to what it's being printed over :- ie. if one plate printed on paper achieves 100% trap (coverage), then when it's printed on top of a previous plate it might have 90% trap because it doesn't stick as well to ink as paper, etc.

The other relates to "choke and spread", used to create plate alignment (registration) tolerance.

Graeme Gill.