7 Replies Latest reply on Jul 5, 2011 6:20 PM by Colin Brougham

    Serious Quality Degradation

    FutureUSAFPilot

      In to Out comparison.png

      Hi all,

       

      Well, there it is.  Screen caps (mag 2x) of what goes into Premiere, and below, what comes out... using the highest Windows Media 9 settings.  I get a similar result with H264

       

      The source is a screen capture of software... and that .avi looks great.  It's the kind of thing you do when you want to make a tutorial video.

       

      I'm stumped.

       

      I see all kinds of good quality stuff out there, so somehow people are producing great quality output.

       

      I'm using two pass, Maximum Render Quality, Decoder Complexity: Auto, Maximum Average Video Bitrate.

       

      Disheartening

       

      Am I asking too much?  I sure hope I'm doing something stupid...

       

      Any suggestions appreciated.

       

      Thank you!

        • 1. Re: Serious Quality Degradation
          Colin Brougham Level 6

          If you're using progressive footage (which you are or should be, since they're screen captures), be sure you're using a progressive (that is, non-interlaced) sequence, and exporting to a progressive output.

           

          Confirm what your footage, sequence, and export settings are--be specific.

          • 2. Re: Serious Quality Degradation
            FutureUSAFPilot Level 1

            Hi Colin,

             

            Thanks for the assist...

             

            My footage is 720x480 @30p.  The pixel aspect is for widescreen (1.2121 due to the video I shot on the Canon XL2, which is DV anamorphic).

             

            Sequence: (all ghosted, unchangable:)

                      Editing mode: DV NTSC 29.97 fps

                      Video: 720x480 16:9, Lower Field First (which sounds like interlaced???)

             

            Export:

            Windows Media,  preset: NTSC Widescreen source to HQ download

            Summary - Output

                      720x480, De-interlaced

                      96 or 10,000 kbps (tried both)

                      Compressed, VBR, 1 and 2 Pass tried

             

                      720x480

                      29.97 fps

                      Pixel aspect: 1.2121

                      Keyframe Interval 5 sec

             

                      Decoder Complexity: Auto

             

            MRQ on

             

            I hope you see something in there...

             

            Thanks again.

            • 3. Re: Serious Quality Degradation
              Colin Brougham Level 6

              So, just to clarify: you're videotaping the computer screen, not running a screen capture utility?

              • 4. Re: Serious Quality Degradation
                FutureUSAFPilot Level 1

                Sorry, I was not clear.

                 

                I have video, shot in DV widescreen...

                 

                AND

                 

                I have screen capture material...

                 

                and I am cutting back and forth between them.

                 

                Live action, capture, live action, capture, lather, rinse, repeat.

                 

                Sorry... I'm a bit new at this... I am beginning to think it was a mistake to get the Canon XL2...

                 

                 

                • 5. Re: Serious Quality Degradation
                  Colin Brougham Level 6

                  OK, I think I'm following...

                   

                  Well, if you did shoot 30p with the XL2, you're in good shape. You just need to make sure that the footage is being interpreted properly as 30p. Right-click a clip, go to Modify > Interpret Footage, and check what the Field Order section is showing. Premiere can be a little goofy with DV at progressive frame rates, so as long as you did actually shoot 30p, there is no harm in setting this to No Fields (Progressive Scan) if it's not already.

                   

                  With that set, you need to decide on a sequence setting. If you're planning on only going to the web with this, you don't necessarily have to use a DV sequence. In fact, you might be better off using a custom sequence with square pixels (DV does not use square pixels, but web video does), but let's keep it simple for now. You can start with--for example--the DV Widescreen sequence, and then change the Field Order to No Fields (Progressive Scan) on the Settings tab. You could also just drag one of your field-corrected DV clips to the New Item button at the bottom of the Project panel, and that will set up a sequence that should be correct.

                   

                  At this point, your source footage is progressive, your sequence is progressive, and you can now export to a progressive output file without worry of deinterlacing issues. Let me know how that goes

                  • 6. Re: Serious Quality Degradation
                    FutureUSAFPilot Level 1

                    Wow, I just learned SO much...

                     

                    You, sir, are now a legend in my mind.  All problems neatly solved.

                     

                    I am curious, however, as to what custom settings I might want to persue for web deployment (escaping the magic that is DV).

                     

                    Might you have some suggestions to a format that would be happy to use my anamorphic DV widescreen?

                     

                    Thank you again.

                    • 7. Re: Serious Quality Degradation
                      Colin Brougham Level 6

                      Glad to help

                       

                      When you're going to the web, you're pretty much unshackled from anything remotely resembling broadcast specs. You can use any frame size and frame rate you want. However, there are some general guidelines you can follow to get the best results.

                       

                      If I'm doing something that is only going to the web, I'll often use a sequence preset that matches the dimensions of the file I want to export to. And, since I'm usually working with a mix of assets, I typically work with the lowest common denominator. In your case, that's the DV footage. It's non-square, as we've established, meaning that the pixels are actually wider than tall--but computers like square pixels, so let's come up with a number to use.

                       

                      The one thing we know is that our height--480 pixels--is constant, so we need to establish the width. Widescreen aspect ratio is typically expressed as 16:9--that is, 16 units wide by 9 units tall, when you're talking about square pixels. There's lots of ways to arrive at the math, but for fun, let's try algebra!

                       

                      16/9 = x/480

                      (16 * 480)/9 = x

                      7680/9 = x

                      853.3333 = x

                       

                      So, that means that our square pixel DV widescreen sequence would be 853.3333 x 480. Sweet! But wait...

                       

                      Fractional pixels really don't jive with video dimensions, so we're OK if we just round that number off. 853 pixels would be closest, but video dimensions really prefer even numbers. So, we could round down to 852 or up to 854, and we'd be OK--you'll see people using all three of these dimensions for such widescreen projects. Personally, I don't use any of them, and go a slightly different route...

                       

                      A lot of video codecs--particularly those used for web deployment--work best when the dimensions are evenly divisible by 16. This is usually called "mod16," and is important because the codecs are usually tuned to work most effectively when blocks of video can be divided into 16x16 blocks, then 8x8 blocks, and then 4x4 blocks (there are rectangular blocks, too, like 8x16). If a dimension is divisible by 16, it's divisible by 8 and 4 as well, meaning that the codec can be most efficient.

                       

                      With that in mind, I try to find the closest dimension that is mod16. In the case of square pixel DV widescreen, that's 848x480:

                       

                      848/16 = 53

                      480/16 - 30

                       

                      Nice, round, evenly divisible dimensions. You lose a few pixels on the side, but that's not a big deal. You can export directly to those dimensions then (or smaller, if you need to), and you're maximizing your encoder's efficiency.

                       

                      Now, in real world practice, you probably get a only the tiniest fraction of an efficiency increase, but--it doesn't hurt to try. You've got a few numbers you can play with, and the general formula to cook up your own.

                       

                      Hope that helps