13 Replies Latest reply on May 5, 2011 3:15 PM by AtonMusic

    How many CS Increments will it take to handle ProRes Correctly

    AtonMusic Level 2

      Two Simpleton-tests with AE 5.5 revealing that AE still hasn't learned a darn thing in regards to the ProRes Codec

       

      Test 1:

       

      1) Import ANY Canon 5DMKII into AE CS5.5

      2) Add it to a new Comp

      3) Add that comp to the render queue

      4) Select ProRes 4444 as output codec

      5) Render and compare with your original Canon Footage

       

      Expected Result in the Quicktime Player 7:

       

      The ProRes File Should look somewhat like the original.

      It does NOT.

       

      Actual Result

      Gamma is WRONG

      Color is slighty shifted

      And Noise/banding has been increased significantly.

       

       

      Test 2:

       

      1) Open Apple's Compressor

      2) Import the same Canon File

      3) Setup a 4444 Codec like in AE

      4) Render out

       

       

      Expected Result in the Quicktime Player 7:

       

      The ProRes File Should look somewhat like the original.

      It does.

       

      Actual Result

      ProRes transcode looks 100% identical to the source

       

       

      Test 2 Expanded

       

      Import the Compressor-created 4444 file into AE and repeat test 1

       

       

      Expected Result in the Quicktime Player 7:

       

      The ProRes File Should look somewhat like the Compressor-Created file.

      It does NOT.

       

      Actual Result

      Gamma is TOTALLY WRONG

      Color is LESS shifted than when exporting the Canon file directly

      Noise/banding is MORE than the compressor created file but LESS than when exporting the canon file directly...

       

       

       

       

      I am more or less speechless. I have been reading about this problem on a million blogs. Including adobe's own.

      Since at least since CS3. There have been several invain work-arounds. And now this. CS 5.5 - with the same problem.

       

       

      Adobe, how many versions do you need to get it right.

       

      Perhaps, you could look a little a Apple's technology, as it seems to be much more effective and advanced than yours.

        • 1. Re: How many CS Increments will it take to handle ProRes Correctly
          Rick Gerard Adobe Community Professional & MVP

          Your color settings are wrong in AE. Make sure that you are either using color management correctly or that you have color management turned off and the composition is set to use legacy gamma.

           

          The attached screenshot shows an After Effects Project titled Gamma Test at the bottom Right, Render Test at the top right. The inset outlined in white is the rendered Animation Codec movie, The outside is the rendered ProRez 444 movie. They are identical in AE. To the left are the quicktime player versions of the Animation Codec and the ProRez render. The Gamma's are identical to the AE file. The Animation Codec renders displays in Quicktime with a slightly different gamma and it does not exactly match the AE Comp Window.

          RenderTest1.png

          This second test shows QT renders from the media encoder rendering the AE project files directly and re-renders of the original 444 render from Compressor. Just for fun I threw in renders to ProRez 422 HQ. Interestingly enough, if you open up these screen grabs in Photoshop you'll find that the ProRez 222 files have slightly different color values for each stripe and that true gray is gone. There is a very slight color cast. The 444 renders are identical, the Animation render is a little brighter especially in the darker colors.

          RenderTest2.png

           

          What does this prove? That ProRez can exactly match AE's composition window when compared as a screenshot. I also opened the QT renders in Photoshop and all of them had exactly the same values for each bar. The color shift in the ProRez 422 files was also gone. I tried the same thing in Premiere Pro. The colors matched. I couldn't open the files in FCP because I don't have it installed on this computer. Turns out that in an adobe workflow the gamma can match exactly. If I get a chance I'll test in FCP tomorrow at work. One more thing. I won't deny that AE does not add a meta tag about color space to the proRez renders but I will claim that the workflow isn't broken. I use it every day.

           

          Just in case you're interested here's the file I used to generate these tests. It may have been more interesting if I had used the Kodak Test image.

          1 person found this helpful
          • 2. Re: How many CS Increments will it take to handle ProRes Correctly
            AtonMusic Level 2

            Hi Rick,

             

            first things FIRST ;-)

             

            Thank you very much for taking the time to make that post.

             

             

            My color settings in AE are correct. And I am using a color managed workflow -)

            It plays NO PART whether I enable the "Use Legacy Gamma"....

             

            Your example is correct in BUT it only applies to the realm in which you created it > RGB

            and it does not exhibit temporal H264 artifacts stemming from a YUV 4:2:0 (609 Color Space) Matrix

            You dont exaclty round-trip the way that I was describing a rountrip in my initial post.

            I know that a ramp can reveal certain things but it is a highly virtual comparison and does not really apply to live-worl  footage.

            At least in this case ;-)

             

            You take RGB-generated grays and then output to ProRes

             

            I took a YUV-color-realm movie entered into the realm of RGB (AE) and then exported to ProRes

            And THAT is where the problem lies.

             

            I am uploading the file in question. It is a movie from the 5DMKII.

            That is the source I have been testing with all along.

             

            It is great in that it has a dark area with quite a bit of noise. THAT noise after AE has encoded it to ProRes will look banded and become worse.

            Not so when encoding with Compressor.

             

            Besides, and this I cannot wrap my head around, when compressor is compressing the exact same file it is 295MB in size but when doing it with AE it is 413.

            Something wrong there ?

             

            Anyway, here is the link to the file if you wanna try with that.

             

            I am uploading a screen shot to show you which area you should look out for.

            And rememeber this is NOT visible when looking at it in AE. you HAVE to rountrip and open the source and the rountripped file in the Quicktime Player.

            Remember, most folks are WATCHING the movies in the quicktime player on OS X... So I would LIKE my files looking identical to the source when watched in the QT player. I dont believe that any consumer out there, just aiming to watch a movie has AE installed so that he can watch the movie in the way I (all AE users) intended it to look.

             

            Adobe we NEED this to work.

             

            When/if you do the test - make a real world comparison.

             

            Open the Original source in AE and out-encode it to PR4444.

            Then open the quicktime player and load up both (PR4444 and ORG source)

            Put them on top of each other and flip back n forth. Huge difference in the area that I have hi-lighted in SS below.

             

            The noise in the original is very present in the original file in that area...

            When looking at adobes encode the noise has become large square fields(almost) looks a little like banding.

            When using Apple's compressor to rountrip it looks exactly like original.

            Might I add, taking the original file and opening it directly in AME (bypassing AE altogether) will have the same impact on the file

            as when exported directly from AE. So it seems as if adobes Quicktime encoding framework has a bug altogether. I tried all of this with the

            cineform codec as well. When using adobe to encode to cineform it displayed the same artifacts.

             

            Original Source Footage Download

             

            Screenshot of area to look for

             

             

            SShot.jpg

            • 3. Re: How many CS Increments will it take to handle ProRes Correctly
              Dan Ramirez Adobe Employee

              Before I address the issue, I'd like to say thanks for caring enough about the product to post your problems here. We appreciate the dialog and the application is better for it. I'd love to dedicate more time to answering questions in the forum, but I'm not tech support; I'm a software tester on the After Effects team and it's hard enough keeping up with our amazing dev engineers as it is.

               

              We've done a ton of work getting gamma and color right with FCP and Apple workflows. You might be shocked to hear this, but the problems are not all Adobe's fault. I've filed a few Apple Radar bugs for issues around this area and let's just say I'm not holding my breath that they'll get fixed. We've had a lot of display problems with QT Player, both 7 and X, particularly with H264 files. I wish there was a common professional video player, but there isn't.


              That said, I'm not seeing the problems you are. I've attached several files for you to check out.

               

              CS55_ProRes4444Gamma folder.zip is an AE project you can use to recreate any of the screenshots below. The two pieces of "missing footage" are your original H264 source file and the ProRes4444 export. Duplicate and render the RQ item to create the ProRes4444 export file.

               

              screenshot_Diff_betweenOriginalandAEExport.png - This image is a diff in AE between your original Canon H264 file and a ProRes4444 export out of AE. This tells me that a ProRes4444 export out of AE matches the color/gamma/quality of the source file almost perfectly.

               

              screenshot_Diff_betweenOriginalinQTXandAEExportinQTX.png - This image is a diff in AE between screenshots of QTX using the original source file and the AE exported ProRes4444 file. It's not a perfect diff, but I'm not seeing a .2 gamma shift or a serious color shift.
              ---

               

              I've done comprehensive roundtrip testing with FCP and ProRes4444 and I haven't seen issues there either. I tried both QT7 and QTX and got identical results, but you might want to try QT7 because there are display gamma differences between the two applications.

               

              I should mention that you'll get inconsistencies in color/gamma with Apple's Color, QT7, QTX, and FCP without even adding AE to the equation.

               

              Last, but not least, I also haven't mentioned that the presence of AJA and other 3rd party QT codecs can affect completely different codecs. To troubleshoot that issue, remove all 3rd party QT codecs from the volumes/harddrive/library/quicktime folder, and see if the problem goes away.

               

              Dan R.
              AE QA Engineer
              1 person found this helpful
              • 4. Re: How many CS Increments will it take to handle ProRes Correctly
                AtonMusic Level 2

                Hi Dan,

                 

                 

                thanks for looking into this. MUCH appreciated.

                 

                 

                Just be clear on MY end...

                 

                If I import the round-tripped material into AE - and compare in AE, the differences. They are not worth mentioning.

                The difference matte I created would have to be increased by 8 exposure stops to make the difference visible.

                 

                The problem, however, isn't so much when comparing inside of AE. The problem is QuickTime and Apple...

                Especially latter.

                 

                The problem is that all my clients (and those of any AE-user having Mac clients) is that they USE quicktime to watch what we deliver.

                They dont use After Effects. If they did, I would not care about how the heck my file looked inside of Quicktime.

                 

                Unfortunately, they do.

                 

                Quicktime Player 7 DOES indeed display files in a different gamma space than that of quicktime 10.

                That is also mentioned by Apple.

                 

                However, the GAMMA I can correct for. But what I cannot correct for is that my noise (in the H264 movie I sent) is turned into square 10x10 pixels

                noise ;-) when encoded via adobe encoders. I get the same weird artifacts when using AME....

                 

                 

                Do you think perhaps that this problem could have something to do with the Importation of file initially. As I stated earlier, if I import an already to-ProRes-Encoded file and THEN export from AE the noise(square)artifacts are not there.

                 

                And LAST but not LEAST ;-) --- A 4444 Compressor or Quicktime Player7 encoded (My H264 File) movie is 295 MB in Size. The exact same file encoded via AE or AME is 412.6 MB....

                 

                If importing the from Compressor-already-encoded 4444 file into AE and then re-encoding it to 4444 the file size is 393.5 MB....

                 

                There must be some kind of weirdness going on here.

                 

                 

                I will try to remove ALL codec but the ProRes from the quicktime folder and try again....

                • 5. Re: How many CS Increments will it take to handle ProRes Correctly
                  Dan Ramirez Adobe Employee

                  I hear you. That's why I attached the diff of the QTX screengrabs. It's not perfect, but I suspect it's a lot better than what you're experiencing. The QTX screengrabs are the original source in QTX and a prores4444 AE export of the original source also viewed in QTX.

                   

                  The H264 decode and ProRes4444 encode issues are likely very nuanced. The size of the ProRes4444 encoded file is probably directly proportional to the data fidelity of the AE comp. I'm guessing that the AE comp with a single Canon H264 file has more color values per px than the same file compressor. These values would be derived from the h264 decode.

                   

                  Bear in mind; I'm not claiming that Adobe's H264 decoding is superior, just that it probably results in more color values per px. If the filesizes issue is a concern of yours; I'd recommend trying to encode a TIFF/DPX/PNG sequence to ProRes4444 in both AE and Compressor. You tried with a ProRes4444 input file and saw that the file sizes were much closer. My guess is that this ProRes4444 encoder issue is actually related to the H264 decode.

                   

                  I don't know exactly what Apple does, but I do know some H264 decoders can introduce a subtle blur which can reduce compression artifacts at the expense of image sharpness.

                   

                  AE and AME don't do anything like this on import, but I could see how one could deem the decoding as inferior.

                   

                  Are you seeing the 10x10 blocks in your AE comp with the Canon H264 as a source? If so, that problem is unrelated to the ProRes4444 export.

                   

                  Dan R.

                  AE QA

                  • 6. Re: How many CS Increments will it take to handle ProRes Correctly
                    AtonMusic Level 2

                    Dan Ramirez wrote:

                     

                    Are you seeing the 10x10 blocks in your AE comp with the Canon H264 as a source? If so, that problem is unrelated to the ProRes4444 export.

                     

                    NO..... ONLY when viewed in the Quicktime Player.

                     

                     

                    Dan,

                     

                    I'd like to take a moment and express my most sincere THANKS for taking the time to write here.

                    Seeing such effort on a public support forum is NOT something I am seeing much these days.

                    And your explanations has certainly put my mind at ease.

                     

                    So thanks a bunch

                    • 7. Re: How many CS Increments will it take to handle ProRes Correctly
                      Dave LaRonde Level 6

                      I recall at the very beginning of this thread that you didn't RENDER out of AE -- using AE's Render Queue -- but you instead EXPORTED.  In AE-Land, the terms "render" and "export" aren't interchangeable.  They involve two separate processes.

                       

                      Because AE's a 64-bit application, but key components of Quicktime are still 32-bit, AE has a tough time if you EXPORT a quicktime.  However, if you render a quicktime, life is good.

                      • 8. Re: How many CS Increments will it take to handle ProRes Correctly
                        AtonMusic Level 2

                        Dave,

                         

                        in all my referrals to EXPORT, I meant RENDER... Or better yet "Add to Render-Queue"

                         

                        And NO - Life is ANYTHING but good WHENEVER quicktime players are involved.

                         

                         

                        One thing I HATE more than Tsunamis hitting coastal areas is Quicktime Players

                        • 9. Re: How many CS Increments will it take to handle ProRes Correctly
                          Dave LaRonde Level 6

                          Well, you must be doing SOME tiny small thing wrong, because I don't have any problems with quicktime.  I wish I knew what it could be.

                          • 10. Re: How many CS Increments will it take to handle ProRes Correctly
                            Dan Ramirez Adobe Employee

                            I'm happy to help. I wish I had the time to help on the forums more often. The person you should really thank is Todd Kopriva for roping me into the ProRes issues

                             

                            Can you post a screenshot of the actual visual artifact in QTX?

                            • 11. Re: How many CS Increments will it take to handle ProRes Correctly
                              AtonMusic Level 2

                              Dan,

                               

                              gimme a few hours. And I'll post you a screen shot.

                               

                               

                              Todd.... THANKS

                               

                              I just tested some VERY nice tools from Digital Rebellion.

                              They have a "Gamma Shift Detector"

                               

                              Below this line is the results coming from them(Perhaps this is something to look at)

                              Perhaps it is NOT your pixels but the FLAGS that your are not setting... I dont know, this is beyond a

                              creative persons logical thinking capabilities

                               

                               

                              ProMediaTools "Gamma Shift Detector" Findings (Destination File = The AE to 4444 encoded file & Source equals 5D H264)

                              __________________________________________________________

                               

                               

                              Checking files...


                              Destination movie's pixels are -3.12% darker


                              Source gamma metadata: 2.20


                              Destination file has no gamma metadata set.


                              Source and destination gamma metadata do not match.


                              Source color metadata is set.


                              Destination file has no color metadata set.


                              Source and destination color metadata do not match.


                              Source color space metadata is set.


                              Destination file has no color space metadata set.


                              Source and destination color space metadata do not match.


                              Recommendation:

                              The gamma shift is occurring within the pixels of the movie and within metadata. You are advised to go back to the application that created the destination file and check the input and output settings to remove the cause of the gamma shift. If necessary, return to Gamma Shift Detector and sync the gamma metadata.

                              • 12. Re: How many CS Increments will it take to handle ProRes Correctly
                                AtonMusic Level 2

                                Dave,

                                 

                                 

                                try downloading the source file I uploaded and follow the steps in my first post.

                                Then look at THAT file (Check my highlighted area) against the original in Quicktime Player 7..

                                 

                                I promise you that there is a difference. A difference that your EYE cannot see if compared in AE.

                                Your eye (if it still works (Sorry just kidding)) CAN see the difference in the QT7 player.

                                 

                                I have tried on 3 different Mac - all showing the exact same behavior

                                 

                                 

                                Best

                                 

                                M.

                                • 13. Re: SOLVED --- SOLVED --- SOLVED ---- YIPIIIIIIIIII
                                  AtonMusic Level 2

                                  Dan,

                                   

                                  the META data is at fault. Or should I say the ones adobe doesn't flag ;-))))

                                   

                                  after writing this post - clicked the Button in the app called "Sync Meta Data"

                                   

                                  ET VOILA.... ALL Square Noise Pixels GONE. And When enable Final Cut Studio Color is checked in QT7

                                  the SOURCE and the AE-Encoded 4444 match 100%....

                                   

                                   

                                  PLEASE LOOK INTO THIS APP AND SEE WHAT THEY DO..... It would BE GREAT

                                   

                                  Then analyzed again with below result (The file changed in view but no enough

                                   

                                   

                                   

                                   

                                  Checking files...

                                   

                                  Destination movie's pixels are 0.39% lighter

                                   

                                  Source gamma metadata: 2.20

                                   

                                  Destination gamma metadata: 2.20

                                   

                                  Source and destination gamma metadata match.

                                   

                                  Source color metadata is set.

                                   

                                  Destination file has no color metadata set.

                                   

                                  Source and destination color metadata do not match.

                                   

                                  Source color space metadata is set.

                                   

                                  Destination color space metadata is set.

                                   

                                  Source and destination color space metadata do not match.

                                   

                                  Recommendation:

                                  The gamma shift is occurring within the pixels of the movie and within metadata. You are advised to go back to the application that created the destination file and check the input and output settings to remove the cause of the gamma shift. If necessary, return to Gamma Shift Detector and sync the gamma metadata.