8 bit vs 16 bit vs 32 bit?

[DrewCCU]

8 bit vs 16 bit vs 32 bit? These are the color mode options in Photoshop. I think I vaguely understand the difference between the 3, but when it comes to digital painting in photoshop, how do you determine which one to use?

[DoorKnobHandle]

That's the color depth, it determines how many different colors your image can have. I don't think, with digital painting, there's much of a reason to go with 32-bit as opposed to 16-bit as you will probably still use a palette when drawing. Photos you take do benefit much more from a higher color depth as they can easily have TONS of totally different colors.

[Snarky]

There are a lot of filters and stuff in Photoshop that only works in 32-bit, and in 16-bit there's also very noticeable banding on gradients (you still get visible banding in 32-bit, but not as bad).

I'd go with 32-bit every time.

[Ghost]

Depends totally on what you want to draw, I'd say. Very fine gradients and most filters look perfect in 32bit and, with luck, decent in 16bit. For digital painting both are valid options, but just to make sure- as Snarky said- go with 32bit.

If you want to use the graphics in AGS, take care your game uses the same colour depth! Importing a 32bit sprite into a 16bit game is prone to mess up your finer gradients. The other way round is always safe- you can import anything into a 32bit game.

Personally I prefer 16bit. Much more flexible than 8bit, and not quite as resource-eating as 32bit. Then again, almost all my stuff is pixel art with no filters applied. I could as well STOP using PS

[Calin Leafshade]

Then again, almost all my stuff is pixel art with no filters applied. I could as well STOP using PS

www.aseprite.org

[InCreator]

In almost 2012 there's no matter. Resources are not a question at all, especially for an AGS game since lousiest of flash drives (1-2GB or so) can easily carry a full game and nobody uses 56K modem anymore.

but to make it clear:

8-bit = 256 colors
-- outdated, but ok for sprite art. I wouldn't limit myself to this palette though.

16-bit = 65,536 colors
-- all you need. Although, If you're heavy gradient user or as said - digital painter, move forward

32-bit = 16,777,216  colors
-- Human eye recognizes around 10 million (impossible to really determine, but popular opinion says so), so it's an overkill. Gradients are smoother though

48-bit = 281.5 trillion colors.
-- yep, we went there and even 64-bit. Not sure why. Cool to know though

[DrewCCU]

In almost 2012 there's no matter. Resources are not a question at all, especially for an AGS game since lousiest of flash drives (1-2GB or so) can easily carry a full game and nobody uses 56K modem anymore.

but to make it clear:

8-bit = 256 colors
-- outdated, but ok for sprite art. I wouldn't limit myself to this palette though.

16-bit = 65,536 colors
-- all you need. Although, If you're heavy gradient user or as said - digital painter, move forward

32-bit = 16,777,216  colors
-- Human eye recognizes around 10 million (impossible to really determine, but popular opinion says so), so it's an overkill. Gradients are smoother though

48-bit = 281.5 trillion colors.
-- yep, we went there and even 64-bit. Not sure why. Cool to know though

Thanks. You guys have been helpful.

[Snarky]

First of all, let me point out something rather important we've all forgotten about (and which makes most of the discussion here moot): The color depth options in Photoshop are per channel!

There are three channels, RGB, so 8 bits/channel is really a 24-bit color space. 32-bit images actually use 24 bits for color and 8 bits for transparency, so 8 bits/channel is the same as a 32-bit image. The 16-bit/channel and 32-bit/channel formats are useful if you're doing a lot of tweaking of brightness/contrast etc. (which tends to increase the difference between similar colors), but since all consumer-level screens only support 24-bit color you're only seeing 24 bits at a time, and the final image is almost certainly going to be reduced back to 8 bits/channel.

So you can safely work in 8 bits/channel: that's the standard Photoshop format and the one everything is compatible with.

16-bit = 65,536 colors
-- all you need. Although, If you're heavy gradient user or as said - digital painter, move forward

You don't have to do a lot with gradients to see limitations in 16-bit color. For example, 16 bits gives you only 32 shades of pure gray (you only have 5 bits in the red and blue channels), which is obviously insufficient to get smooth transitions without dithering. And there's no transparency, either.

In fact...

Very fine gradients and most filters look perfect in 32bit and, with luck, decent in 16bit.

Eh, try drawing a black-to-white gradient over an 800x600 canvas in 32-bit, and take a look at the darker part. You'll be able to see very obvious, ugly banding between the shades.

32-bit = 16,777,216  colors
-- Human eye recognizes around 10 million (impossible to really determine, but popular opinion says so), so it's an overkill. Gradients are smoother though

I see that various references say this, but I find that you can generally see a difference between two adjacent 32-bit RGB values (though you'll have to look closely in many cases), and since the gamut of human vision (the range of colors we can perceive) is much greater than what can be displayed on-screen, we must be able to differentiate more colors than that.

The other point is that the RGB color space does not map perfectly on to human vision, particularly when it comes to brightness: human perception of brightness is logarithmic while the standard computer representation is linear. That's led to the development of HDR formats for photography, and the use of floating-point representations for colors instead of integers.

And yeah, finally there's the issue of having extra accuracy for things like brightness/contrast adjustments, gamma correction etc.

[Ghost]

The other point is that the RGB color space does not map perfectly on to human vision, particularly when it comes to brightness: human perception of brightness is logarithmic while the standard computer representation is linear.

Not to hijack the thread, but what does that mean- sounds valuable to know.

[Snarky]

It means that we distinguish differences in brightness on a relative scale rather than on an absolute scale. So, for example, we might be able to distinguish two light sources where B is 1% brighter than A. If the brightnesses are low (say 1 and 1.01, on some scale of light intensity), that's a small absolute difference (only 0.01), while if they are high (say 10'000 and 10'100) it's a large absolute difference (100), but to us they'll appear about the same. Or in other words, our eyes adapt their sensitivity to the amount of light.

[Anian]

Not to mention if you go into print, then there's a whole bunch of problems on top of that.

Not to hijack the thread, but what does that mean- sounds valuable to know.

It's a function thing, you probably know the difference from a math class.
I tried to explain, but then the thing got to long so here http://www.covingtoninnovations.com/dslr/Curves.html (it's near the begining)

[InCreator]

I work in printing and for what I've seen, blue and black tones are always most problematic for CMYK printing. Even with perfectly chosen Pantone color codes, totally sane clients, well calibrated monitor and good eyesight for designer, expectations never meet the outcome. For some reasons, reflected light has less problems than (how do you say this in english -) absorbed light, so darker tones have more problems.

In given diagram, CMYK are is outside on lighter color side, which is actually less error prone.
But we print for advertising anyway and advertisements are for ordinary people who rarely even realize advertisements, because we're used to ignore. So aside picky client, target group never realizes that say, coca-cola red isn't really same red on bottles and that big poster.

[Anian]

But don't you get the "Coke red" color special?
I know for example print of National Geographic here uses a fifth additional color for that border gold/yellow, same goes for T-Com/Mobile pink I think (though I only saw the NG printing). Esepcially for a "brand" that is really recognizable, such as Coke, I mean the poster colors are more UV and water proof, but still aren't there like outlines for stuff like that, mixture recepies that have to be followed?
But in the end, that's why you do a test print for the client to sign off on.