I have recently gotten a tablet to draw on using Krita and used the following video to learn an easier way to complex shading from a professional. I know he is using photoshop, but Krita's manual states the functions are the same.
https://www.youtube.com/watch?v=vZ5ulFxA-eA&t=515s

So, using the tutorial, I coloured my character[1], created a shading layer of a soft pale blue and a highlight layer of soft pale yellow[2], and when I went to apply the respective multiply and overlay filters, and something odd happened[3]...

https://drive.google.com/file/d/11ChBw6 ... sp=sharing [1]
https://drive.google.com/file/d/1P5_KG3 ... sp=sharing [2]
https://drive.google.com/file/d/181RhQq ... sp=sharing [3]

I don't know what happened... please help.

P.S. the splotches on the side are the desired highlights and shading of the face.... pulled from a reference picture

If a contributor here is familiar with these techniques, it would be very useful to provide a link to download your .kra image file so it can be examined for possible problems or misunderstandings.

EDIT: P.S. Using the same methods as shown in the linked YouTube video (but with much less skill and ability), I've made shadow and highlight layers over the original image that you used:

https://imgur.com/JHXi4NP

This seems to work. I don't know much about blending modes but it may be a good idea to try the various Lighten and Darken blending modes that are available in krita. The manual gives a good description of how they work and what they do.

https://docs.krita.org/en/reference_man ... t=blending

I dunno where I went wrong..

here is the file

https://drive.google.com/file/d/1BgypCP ... sp=sharing

Thank you for providing the .kra file, that's always useful.

You have all layers set to Normal but I think that that Layer 48 is shadows so I set it to Multiply blending mode and set Layer 46 to Overlay mode since it seems to be the desired highlights. The results were not good.

However, you're using the CMYK/Alpha colour space. I assume this is because you intend to print the image. I think the general advice is that you should use a standard RGB colour space and then use the 'proofing' facility to see how it will look when printed. Someone else may be able to give you better advice on that.

I converted the entire image to RGB 8-bit colour space (Image -> Convert Image Colour Space, which took a long time) and the thing now seems to look ok but the highlights are too intense. The details of which colour space to use in which situation are a bit beyond me but someone else may be able to give advice about that.

As a matter of interest, you can use Overlay mode for both shadows and highlights by using dark grey for shadows and light grey for highlights with a 50% opacity brush. Vary the level of grey and the brush opacity for different effects. This may not be the best way of doing it but it seems to work.

:I think the general advice is that you should use a standard RGB colour space and then use the 'proofing' facility to see how it will look when printed. Someone else may be able to give you better advice on that."

Yes, that's 100% correct. Always work in RGB, then convert to CMYK if your printer needs it. Not all printers do; it's something to check before you start working. If they do, get their CMYK profile and set it as a proofing transform. https://docs.krita.org/en/general_conce ... kflow.html might be interesting for Victor.

ahabgreybeard wrote:You have all layers set to Normal but I think that that Layer 48 is shadows so I set it to Multiply blending mode and set Layer 46 to Overlay mode since it seems to be the desired highlights. The results were not good.

However, you're using the CMYK/Alpha colour space. I assume this is because you intend to print the image. I think the general advice is that you should use a standard RGB colour space and then use the 'proofing' facility to see how it will look when printed. Someone else may be able to give you better advice on that.

I set the layers 46 and 48 to normal for safety sake, hoping to avoid ruining my progress.

as for colour space, I seem only to have alphas and chose CMYK/Alpha since I was under the impression that it was more detailed than RGB.
However, I did try RGB/Alpha and got a brighter version of the same problem... Where do I find non-Alpha colour spaces?


I am new to Krita...

The number of pixels is what makes for detail, not the colorspace. Some colorspaces can show more colors than others: cmyk has fewer colors than rgb. If you're new, it would be a really good idea to browse the manual a bit. There's lots of getting-started information there.

boudewijn wrote:The number of pixels is what makes for detail, not the colorspace. Some colorspaces can show more colors than others: cmyk has fewer colors than rgb. If you're new, it would be a really good idea to browse the manual a bit. There's lots of getting-started information there.

I would like to thank you for your help. My knowledge is limited and I have been looking into the manual, however I have been taught photoshop in a more hands-on approach. So I am a little confused on how to apply what I learn....

So far, I don't mind switching to RGB, but all the colour channels I have are all ending with /Alpha and I am concerned that is the cause of the issue I am having... How can I avoid this?

'Alpha' is a measure of the opacity of a pixel. Alpha=1.0 is full opacity, Alpha=0.0 is full transparency.

RGB/Alpha means that the image uses RGB colour space and can present/encode opacity/transparency. Some older or more 'simplistic' painting applications could not do transparency so they are said to use RGB colour space. (I suppose these applications are quite rare nowadays.)

A .jpg image can't have transparency in it but a .png image can. etc, etc.

"I did try RGB/Alpha and got a brighter version of the same problem":

After I converted your .kra image to RGB/Alpha 8-bit, it seemed fine but with intense highlights. I suggest that you try reducing the opacity of the highlights layer to reduce the effect. If you painted the shadows and highlights while in CMYK colour space then maybe this affected your perception of what was 'enough' highlighting or shadowing.

Also, you were using 32-bit floating precision which may have a bearing on all this. 32-bit float precision seems to be overkill for a simple cartoon-like image. Did you have a particular reason for doing this?

If you're just starting and learning about digital image creating and manipulation, it's best to stick with simple and standard techniques in the early stages, like using RGB(/Alpha) 8-bit integer for everything. What you see on your computer screen is in fact always an RGB 8-bit image and the use of higher bit depths in imaging applications is a specialist technique.

ahabgreybeard wrote:Also, you were using 32-bit floating precision which may have a bearing on all this. 32-bit float precision seems to be overkill for a simple cartoon-like image. Did you have a particular reason for doing this?

If you're just starting and learning about digital image creating and manipulation, it's best to stick with simple and standard techniques in the early stages, like using RGB(/Alpha) 8-bit integer for everything. What you see on your computer screen is in fact always an RGB 8-bit image and the use of higher bit depths in imaging applications is a specialist technique.

Digital art in both vector and bitmap methods isn't new to me, but Krita is. As for why I went with 32 bit is mostly from past experiences with other programs. When I worked with 8 or 16 bit imaging, the color options seemed to be limited and I couldn't get a proper gradient like I was searching for. My shading style has often relied on suddle nuances in hopes to give even my cartoon characters a more realistic look.

However, now that my programs were outdated, I searched to find Krita, which is mostly the samr as what I am used to, but with enough differences to merrit a learning curb. I am also a visual learner, so using the manual to learn how the differences work has shown to be a bit more difficult, which lead me to finding a community support forum to ask more specific questions.

As I said, I tried converting it to RGB and even turned the opacity of the highlight layer down (since it lost it's yellow tinge) and I got similar results to the messed up picture I sent. That has an added bonus of reducing the gradients from gradual and fragmented opacity to purple ovals with harsh edges. I don't see how adding even more limitations will solve the problems I am having as it seems to relate to the filters and not the bit rate.

As for screen rates, the reason we use hexidecimal image codongs (hex codes) for images, is because the average screen runs on 16 bits, not the 8 bits that the peimative models were used to.

I appologize if I seem rude, however I have read the suggested manual entries, searched out additional informations, and have done a lot of research even before comming here, however I am left unsure at what information I am missing to fix the problem I am currently facing. I am agrivated at myself for being in a position of not knowing what I don't know, even after all the experience and knowledge I have optained, and am eager to be enlightenend not only on how to fix this problem, but why it happened and how to avoid it in the future.

I know this does not forgive rudeness, and for that I am truely sorry

Oh, wait -- is your image 32 bits per channel? (I cannot download stuff right now, I've been on a train using my phone for internet all day.) That's a really specialized things, mostly for people working on movie frames.

'8-bit integer' refers to 8 bits (value 0-255) for each colour channel which for the RGB representation is a total of 24 bits to represent each pixel on the screen. (The Alpha channel also has its own 8-bit representation within the image processing application.)

The reason for using 16-bit integer (per channel) or even more precision within image processing applications is, as you have found out, to enable cumulative mathematical/arithmetic operations on multiple layers of images while not losing the numerical precision/resolution required to enable things like subtle shading effects, especially for shadow and highlight details. If you used 8-bit integer for these types of operations then you might get 'banding' effects due to lack of numerical value resolution. This is effectively a form of rounding error that can be solved by using higher bit depths to represent an image. This would be especially true for multilayer images with many different blending modes operating throughout the layer stack.

If you find that use of higher bit depths gives you better results in your work then be all means use it.

Here is a screenshot of the Alexa image after I converted to 8-bit integer and set the 'highlight' layer (46) to 50% opacity. I also applied a 40 pixel Gaussian blur to the highlight and shadow layers to smooth the edges.

https://imgur.com/rVKmTXx

I see no significant flaws in the result. I also tried conversion of the original image to RGB/Alpha 32-bit float and there was no obvious difference between the end result of that and the result after 8-bit integer conversion.

Oh, and sorry for the patchy replies -- gradients are, unfortunately, calculated in 8 bit rgb in Krita no matter the image colorspace. It's on my todo to change that, but I haven't started on that job yet.

Thank you Ahabgreybeard and Boudewijn. reducing the bits seems to have fixed most of it... although the face still needed some resorting for the shading...