Announcement

Collapse
No announcement yet.

Advanced Bit Depth Question

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Advanced Bit Depth Question

    I have an incredibly complex image (series) that is killing my machine. 13x18, ProPhoto, 16-Bit Depth with several Smart Objects. While I want to stab myself repeatedly with an exacto knife (changes are slow as molasses), i was hoping for a better solution.

    We know that 16-bit gives the best tonal range, thats why we use it, but I was curious what the group thought about working in 8-bit. No no, wait, hear me out on this one.

    If the base image is a Smart Object still attached to the RAW file, and everything else is an adjustment layer, then (I may be wrong here, but) if the raw is 1's and 0's, and the adjustments are 1's and 0's, then (other than the masks), I *think* I could work in 8-bit, then turn it back to 16-bit after I did all the hard work.

    I mean, if nothing is "technically" pixels (except a few masks), then should I still be worried about banding now? I have not done any cloning yet, because I am still working on the compositing, color and lighting stage.

    These are 4gb beasts, and any ideas would be awesome.

  • #2
    Re: Advanced Bit Depth Question

    If you move to 8-bit you permanently lose all extra information in all your adjustments. That lost information cannot be added back afterward.

    That said, except for banding it can often be impossible to distinguish 8-bit from 16-bit in real life applications. So it might work for you just fine and still provide faster speed and smaller file size. Depends on the image you're working with, how much you've done and what you have left to do.

    One idea could be to save a copy in 8-bit, work on it and see if you see any visible degradation. If not, you're good. If so, then you'll still have your partially finished 16-bit copy to go back to.

    Comment


    • #3
      Re: Advanced Bit Depth Question

      But thats just it... isn't the Curves Adjustment 0's and 1's, and not care if the image is 8 or 16? And isn't the Raw Smart Layer always going to be 10-12 bit? If I go up to 16 later, I would think it would apply that info at that point.

      FYI - I realized that I had CS6 Auto Save feature set for 5 minutes, because of my other easy to use stuff. Basically, with a 4gb file, this thing constantly saving. Once I extended it to 10 minutes (likely go to 15), it became much more usable.

      Comment


      • #4
        Re: Advanced Bit Depth Question

        High bit files are high bit files but not all the feedback from Photoshop is High bit. Perfect example is the Histogram. There's no display I know of that would allow Photoshop to show 64000 odd levels, one per level: the Histogram Palette would need a HUGE display to show you the values.

        What's really clogging you up are the Smart Objects. Each one is the full resolution of the original you've embedded into a big Photoshop doc. If you can reduce them, you'll lower the size significantly while keeping that useful high bit data.

        High bit files don't necessarily have more tonal data per se. They have more steps throughout the tonal scale. Having a 16-bit and 8-bit per color cousin, the scale is the same. The steps between are vastly different. Just as two 10 foot long stair cases is just that, 10 feet. But one can have 10 steps (a foot apart), the other can have 20 (half a foot apart).

        Comment


        • #5
          Re: Advanced Bit Depth Question

          Originally posted by artofretouching View Post
          But thats just it... isn't the Curves Adjustment 0's and 1's, and not care if the image is 8 or 16? And isn't the Raw Smart Layer always going to be 10-12 bit? If I go up to 16 later, I would think it would apply that info at that point.
          Actually, no, for a couple of reasons:


          1. Regarding 0's and 1's, every time you add a bit you double the potential information in the file. Basically:
          • A 1-bit file would have 2 possible levels of color for each pixel
          • A 2-bit file would have 4 possible levels of color...
          • A 3-bit file would have 8 possible levels of color...
          • .....
          • An 8-bit file has 256 possible levels of color
          • A 16-bit file has 65,536 possible levels of color at each pixel
          So the *theoretical* difference between 8-bit and 16-bit is huge. That's why you can get banding with 8-bit files but seldom with 16-bit files. The 16-bit files can make much finer gradations between one pixel and its neighbor, which make smoother tonal transitions.

          2. Loss of data/information

          Once you chop those 8 bits off, you cannot get them back. Every layer gets affected and Photoshop keeps no memory of what the file used to be once you exit the file or do so many steps you lose the 16-bit History. After that History is gone you're 8-bit even if you then convert back to 16-bit.

          Bit depth is like a pot. Color is like a turkey.

          A 16-liter/gallon pot can hold a bigger turkey. But if you then chop the turkey in half so it fits into an 8 liter/gallon pot and give away the other half to your neighbor who eats it, then putting that remaining turkey-half back into the 16 liter/gallon pot doesn't give you back the rest of the meat that you gave away.

          You have more room (aka flexibility) for fancier, more refined dressing, gravy, veggies and other fixings (aka post-processing) than the 8-liter pot would give you but the other half of that turkey is gone forever.

          So the best thing to do is to do the more refined processing in 16-bit then move to 8-bit once you're pretty sure you're finished or close to finished.

          Originally posted by artofretouching View Post
          FYI - I realized that I had CS6 Auto Save feature set for 5 minutes, because of my other easy to use stuff. Basically, with a 4gb file, this thing constantly saving. Once I extended it to 10 minutes (likely go to 15), it became much more usable.
          Great! Hope that solves your problem!

          Comment


          • #6
            Re: Advanced Bit Depth Question

            For what it's worth, the other thing I noticed is that since I had reached a size cap with 4gb TIF file, I tried PSB, which made a smaller file (by 1gb), so I started using that. But, apparently, it's using more processing power (at this size) to crunch the data, and causing more slow down. The TIF saves noticeably faster. On my existing image, PSB took an extra 40 seconds to save, and this is a smaller file than the two I opened this post about.


            So, the morale of this thread, is not so much that 16-bit was my problem, but AutoSave, and PSB. Once I found the issues, I can work normally again, even with many different Smart Objects (which, I actually need).

            Comment


            • #7
              Re: Advanced Bit Depth Question

              Originally posted by artofretouching View Post
              I have an incredibly complex image (series) that is killing my machine. 13x18, ProPhoto, 16-Bit Depth with several Smart Objects. While I want to stab myself repeatedly with an exacto knife (changes are slow as molasses), i was hoping for a better solution.

              We know that 16-bit gives the best tonal range, thats why we use it, but I was curious what the group thought about working in 8-bit. No no, wait, hear me out on this one.
              People worked with 8 bit images for years that looked better than much of what you see today. It's nonsensical to treat the lower bit depth as the boogie man. The only issue here is you're using a very wide gamut, so the distribution of values could run into problems faster due to how thin they're spread at 8bpc. There is no greater range at 16 though unless you're referring to floating point formats that can store values beyond 1, even though they would be automatically clamped if you went down in bit depth or tried to print such a thing. It doesn't apply in photoshop anyway as half-float formats open as 32 bit.

              In your situation youou could simply see if everything fits in a narrower gamut without single channel clipping (relative colorimetric + gamut warning). If your machine can't take it, stop using smart objects. Just reprocess the image again if it really need it. Otherwise resume what you're doing. I can't see why anyone would put themselves through this.

              Comment

              Loading...
              Working...
              X