anchev wrote:
Add a curve layer on top and use it to increase contrast in the zone you are working. Then you will see better the unevenness which will help you to achieve smoother result.

And btw if you are working on a layer in soft light mode you don't need to fill it with gray. The extra pixel data you add by filling it is completely unnecessary and will only increase your file size. Work on an empty layer, that's enough and also more controllable.

Thanks for fast response.  I'm working on an empty 50% gray layer.
I did not ask my question clearly : AFTER making the skin even with some D&B, now I need to "contour" with some brush strokes. I know that my brush strokes will cause unevenness, don't know if lowering flow to 1% will give me satisfaction, and that will increase tremendous work . So I wonder if there's a quick way to achieve gradual highlight-to-shadow smoothness. Maybe I can experiment gradient tool on 50% gray layer?

findart wrote:
Thanks for fast response.  I'm working on an empty 50% gray layer.

You are welcome but as I explained that is inefficient.

I did not ask my question clearly : AFTER making the skin even with some D&B, now I need to "contour" with some brush strokes. I know that my brush strokes will cause unevenness, don't know if lowering flow to 1% will give me satisfaction, and that will increase tremendous work . So I wonder if there's a quick way to achieve gradual highlight-to-shadow smoothness. Maybe I can experiment gradient tool on 50% gray layer?

If what you are trying to is to increase contrast you can do that with curve+mask. Or if you are trying to paint certain detail to create the impression of more relief you can do it the same way you do D&B - use an empty layer in a relevant mode and paint on it. If it looks jaggy/uneven - blur it and mask/delete the extra. Or use a pressure sensitive tablet and tune your brush, soft edge etc.

You can also use an adjustment layer (e.g. to darken or lighten) and paint in the mask and you can fine tune the feathering of the mask.

anchev wrote:
If what you are trying to is to increase contrast you can do that with curve+mask. Or if you are trying to paint certain detail to create the impression of more relief you can do it the same way you do D&B - use an empty layer in a relevant mode and paint on it. If it looks jaggy/uneven - blur it and mask/delete the extra. Or use a pressure sensitive tablet and tune your brush, soft edge etc.

You can also use an adjustment layer (e.g. to darken or lighten) and paint in the mask and you can fine tune the feathering of the mask.

Great! I'll try these techniques one by one.
Thank you so much!

Kevin Connery wrote:
FWIW, I used to use empty layers, but found the 50% gray made it easier to see what I had done. It DOES add more filesize, though.

Here is a tip to save on filesize and still be able to work on a empty layer:

Use a solid color adjustment layer with 50% gray under the layer you paint on. Don't put any mask on it (mask = raster data = size increase). Once you are done with D&B you can delete it. Solid color adjustment layers are much more memory efficient than raster layers as they are procedural.

TMA Photo and Training wrote:
You can often use a curve adjustment layer + mask  with one curve "pulled up" to paint in the highlights where ever you want them... and however intense... or however soft you want it.  The adjustment layer + mask Dodge and Burn technique also allows you to change your mind or modify your work easily.  Paint on the low-light contours with another adjustment layer with the curve "pulled down" for the darker / shadow contours.   Use Control I to invert your white layer mask to become a black hide all mask to begin.  Then use a White brush to add in the lightening or darkening effect...and use a black brush on that same layer mask to erase the effect.  The opacity of your brush should be in the 3-6% opacity range with 100% flow and with a soft edge.  This will make it so that no stroking or overlapping lines show up in your final edit.  You can always gaussian blur the masking to further spread and soften the edges of the applied visual effects.   You can also try using the lighten and the multiply and the luminance blending modes on your layers to get another effect you might like to experiment with.  The overall file size during this phase doesnt usually bother me too much... especially if you have a fairly decent retouching setup.  If you find that your pore level retouching is changing the color of the skin or leaving dark magenta smudges or artifacting...you can also simultaneously use the curve channels (RG and B to) create an applied color correction to the mis-colored pixels.  No 50% gray layer is necessary in the curve based D+B.

Sometimes also, the smoothness or transition of the skin is affected by the brush size, or the brushes softness, and at other times with your own realization that you should begin shading up or down a little bit more or less as you approach a light or dark transition.  Sometimes you might be working at a 1 pixel brush size level...and at some other times you might softly apply a slightly softer edge, or maybe even a much larger size brush size...  to affect not just individual skin pixels...but to affect larger areas of skin that are beginning to transition into lighter or darker regions.  That is called a "transition zone" and if it were to still be done with just single pixel adjustments...that could be the harshness you might be seeing possibly.  There are individual skin pixels you need to pay attention to, skin that makes up a larger pore, skin that is responsible for the average color of that area...and also larger maybe 6 pixel brushes that are soft...that are applied mildly after the previous steps...because the luminosity or color of skin is softly beginning to transition... over some distance... into more of a highlight or a shadow.  That transition is soft and gradual over some distance...and if done with a 1 pixel brush... could end up looking edgy, blotchy, and not smooth.  (It takes a long practiced time to learn the nuances of D+B...and there arent many tutorials that deal with those aspects.)

To further enhance your skills as a retoucher...go to YouTube and type in the search term "Facial Contouring Makeup".  From that search you will see and hear how the makeup artists actually plan for and apply the contouring makeup effects in real life.  They will show you where to apply contouring, how much, and what colors are best to use.    Its really great to see the classy / glamorous difference good facial contouring makes to the final image!

A thought on some terms:  The pixel-to-pixel level dodge and burning you are already doing is sometimes called by some  "Micro Level D+B"  for the small microscopic single pixel areas you are affecting on each pore...  and the larger "Macro Level D+B" you are asking about... is putting back in the larger overall highlights, and the shadows, and the 3 dimentionality, to the broader whole face areas.  The Macro... or overall part of the face retouching... is also sometimes called Contouring in the makeup world.   Not everyone uses this nomenclature... but I find it to be descriptive and enjoy using it.

Wish you the best!

Wow, thank you so much for detailed instruction, that's sweet.
I just experimented this: use a blank 50% gray layer, then add several blank layers on top, each one as a clipping mask layer to that gray layer, and each one deals with one area of the face, thus I can treat different situation like with blur to make it blends smooth.
Or, for example, on one of these blank layers I lasso the right half of the forehead or the whole forehead, it depends, then use a soft brush as big as the whole forehead, click 2~5 times, thus I can get a gradual  "Macro Level D+B" effect.

anchev wrote:
Here is a tip to save on filesize and still be able to work on a empty layer:

Use a solid color adjustment layer with 50% gray under the layer you paint on. Don't put any mask on it (mask = raster data = size increase). Once you are done with D&B you can delete it. Solid color adjustment layers are much more memory efficient than raster layers as they are procedural.

Thank you! I've used color layers for many other things, but never thought of it for this.

anchev wrote:
Use a solid color adjustment layer with 50% gray under the layer you paint on. Don't put any mask on it (mask = raster data = size increase). Once you are done with D&B you can delete it. Solid color adjustment layers are much more memory efficient than raster layers as they are procedural.

How much does one normal layer filled with 50% grey increase the file size? (Or a hundred of them, for that matter?)

Or is this me trolling again?

anchev wrote:
And btw if you are working on a layer in soft light mode you don't need to fill it with gray. The extra pixel data you add by filling it is completely unnecessary and will only increase your file size. Work on an empty layer, that's enough and also more controllable.

I also tested this statement, and the results I get suggest that performing D&B on an empty layer seems to actually increase the file size MORE than doing D&B on a 50% grey layer. I don't understand why this should be, and I don't have time to test this further. Perhaps I've overlooked something.

But the basic principle is that when you fill a layer with one single colour, whether it is grey, red or any other solid colour, photoshop expresses this using only a few bits of data, i.e. all pixels = same value.

Personally I've never bothered with the 50% grey approach - can't see the point.

I think there is an argument that working on a single Curves layer mask takes up slightly less data than working directly on a single D&B layer, but you need two Curves adjustment layers for this approach - a Dark one and a Light one, which cancels out any file size benefit.

But as I suggested before, the difference in file sizes, whichever method you choose, is really insignificant. The important thing is just to use an approach that is effective and comfortable to work with.

RE the OP's observation regarding achieving smoothness. If I understand correctly the issue you're describing, I'd suggest not using Flow at all, and just building up your D&B with repeated 100% soft brush strokes at a very low opacity (e.g. 3 - 5%) This gives you full control of the gradual build-up of the D&B effect.

a k mac wrote:
How much does one normal layer filled with 50% grey increase the file size? (Or a hundred of them, for that matter?)

Test procedure:

CASE 1 (raster 50% gray)

1. File > New, square



2. Simulate real image: Fill background with (128, 128, 128) and add 10% noise:



3. New 50% gray layer (no mask)



4. Simulate D&B:

a) select all
b) transform selection 50% width and 50% height
c) new Gradient layer:

[img]https://snag.gy/AFmLi6.jpg[img]

5. Rasterize layer
6. Apply mask
7. Merge down with the 50% gray layer

The result should look like this:


8. Save as TIFF without compression




CASE 2 (solid color adjustment layer, 50% gray)

All steps the same, the only difference is:

3. New solid color adjustment layer (128, 128, 128), mode: soft light, no mask
7. Change mode to soft light

Results at different resolutions (1k and 5k squares) in bytes:

1k-colorfill.tif - 14021356
1k-raster.tif - 22022160 (x1.57 bigger)
5k-colorfill.tif - 350020620
5k-raster.tif - 500020564 (x1.43 bigger)

And when saved with ZIP image and layer compression:

zip-5k-colorfill.tif - 294222340
zip-5k-raster.tif - 294509256 (less than 0.1% bigger, practically the same)


It makes sense that when saved with ZIP compression the files will be much smaller because the compression algorithm reduces it. However disk usage is not the same as memory efficiency in the sense - amount of actual data which the CPU must process. Also saving compressed files (TIFF or PSD) is much slower than saving non-compressed TIFF (which I have found to work fastest) because compressing the data takes much more CPU time than what the file write time. That's why it is inefficient to use compression during working and saving all the time, only for the final result.

In any case the gray layer is not necessary at all, so there is no need to keep any gray data which does nothing.


The disadvantage of using 2 curve layers is that one needs to switch between the two all the time which is slower than a key press for toggle bg/fg color. Also 2 layers means 2 masks and 2 adjustments on top of each other which I haven't tested in depth in terms of memory efficiency. Curves additionally introduce the potential issue of dodging and burning on top of each other which means one can have raster data in both masks without any effect as one counteracts the other - and that is not efficient too. Plus curves don't simply change the brightness but also the contrast (an hence saturation) in a non-linear way which means more adjustments are needed with more layers etc. The same applies to soft light mode. But I won't go into all that, too complicated and off-topic for current discussion.

a k mac wrote:
@ Anchev

Why are you adding noise. By doing so are you not introducing a massive amount of data that a solid colour layer would not have.

That is just a way to simulate some image data in the background layer by adding randomness. It is only in the background, not in the correction layers. Otherwise it is too synthetic and other compression or procedural factors may influence the result. ETA: factors which are internal to the software, not an explicit function which the user can control. I am looking at it from programmers viewpoint. If you have ever written some code you may know that algorithms can handle uniform data differently from random data.

Your issue with adding 100 layers and ending up with the same file size is most probably because you don't turn off compression.

a k mac wrote:
In normal everyday practice I work with files up to just over 1GB and I've never found solid colour layers of any real significance re. file sizes. But I wonder if someone can explain the discrepancy in my experiments.

Think about memory efficiency in the sense of number of computations. То clarify think of it as 2 arrays of integer numbers A and B, each one with 1 million elements for example.

Computing:
A(1) + 5
A(2) + 5
...
A(1m) + 5

is way faster than:

A(1) + B(1)
A(2) + B(2)
...
A(1m) + B(1m)

a k mac wrote:
Thanks for explaining that. You obviously have an excellent grasp of the maths involved.

Unfortunately I squandered away my maths lessons drawing pictures. But here are some basic 'real world' operations.

1/         250MB, 1 layer document.            Save time    under a second
2/         Add to above 50 grey layers         Save time    under a second
3/         Add to above 5 noise layers         Save time    3.75 seconds
4/         Remove the  50 grey layers          Save time    3.75 seconds

Using surface blur which is notoriously slow, took exactly the same time with or without 50 extra grey layers.

These operations were all performed on an old 2011 Macbook Pro (not exactly the fastest!)

So really I have to ask, what's this big deal about avoiding using one or two grey layers.

If you'll excuse this analogy - How many olympic athletes have totally hairless bodies, despite the indisputable fact that shaving reduces weight?

I don't know if it is a big deal or not. I am just mentioning that it is more memory efficient to work with less raster data. That surely depends on image resolution as well as the factors I mention below. Obviously for 1*1px image it will be the same.

In each computer system there is a bottleneck for a particular type of operation. It can be the CPU, RAM, GPU, HDD, the software algorithm etc. You cannot remove the bottleneck, you can only move it from one node of the system to another. In that same sense: when constructing a test you should make sure you are testing the right bottleneck without the others influencing. Example:

When writing 10Mb of data to HDD, the time for that is very short because the hardware cache of your HDD is 64Mb (for example) and writing to cache is much faster than writing to HDD plates. So you are not testing actual write to HDD. To test actual write, the amount of data should be much bigger than that (e.g. 2GB).

Similarly when testing file format efficiency in terms of time for save, we should isolate the influence of other factors, such as:

- storage speed
- other programs/services running in the background and using system resources
- Photoshop's settings
- OS settings (swapping/caching, driver settings...)
- and so on

For example when I work (and test) I always save to ramdisk as it is more than 10 times faster than popular SSD and doesn't wear. Additionally I have all swapping disabled and my Photoshop scratch is also on ramdisk. This gives me write speed of about 8Gb/s compared to what a normal HDD can do (100mb/s). In this way and by turning all other influence (I even disable network) I isolate the bottleneck entirely on the save algorithm.

So when comparing layered file formats some time ago I tested big TIFF, PSD and PSB with actual data with/without compression and with/without compatibility enabled for PSD. The fastest saves were with uncompressed TIFF.

That said: I can't say much about your test resuls because it is not clear to me what is 250MB - is it the file size, of what format, with what settings and system config etc. Remember also that adding more layers means allocating more memory. Unless you are applying surface blur to multiple layers in a sequence and/or hitting any swapping limit it should not influence at all the time for a surface blur on a single layer. So when testing you may want to consider the explanations so far. Just remember that you must fix all other parameters and benchmark only one.

As for athletes shaving their bodies - AFAIK it is not done for weight loss but rather for different reasons, e.g. to reduce water friction for swimmers.