8 Replies Latest reply on Nov 14, 2012 9:10 AM by Rick Gerard

    ProRes VS DNxHD

    Pierre Devereux Level 2

      Hi,

       

      I have asked previously around the difference between the two codecs, but have not had the oppertunity to use both untill now. The Samurai recording device we use has just received  the latest firmware update, and it now supports DNxHD as well as ProRes. To date, I have been using the ProRes HQ format that the device allows - this gives a 10bit 22MB per second video recording (176mbps) - I shot a 10second clip today using the DNxHD codec, and see that it gives a 10bit 23MBper second video recording (184mbps). So the difference in file size and data rate is rather small (this was using DNxHD 220x codec). I installed the codec on my PC and am able to import both files. I do see a visible difference betrween the two though. The video recorded with DNxHD seems more "washed out" than ProRes. Ill add a screenshot below. both were shot under the same lighting - in fact the only time difference between the two, was the 10 seconds or so it took to change the device from ProRes to DNxHD.

       

      I had the character slightly under exposed so that I could get the aperture a little smaller and more of the image in focus.

       

      ProRes vs DNxHD.JPG

       

      Is this the normal behaviour of the DNxHD codec? Since I have been working with After Effects, I have not run into any issues using ProRes, I am going to assume, untill I am corrected by the forum, that it seems to be the best codec for our current workflow for now

       

      Thank you

       

      Pierre Devereux

        • 1. Re: ProRes VS DNxHD
          Mylenium Most Valuable Participant

          Internally all Avid CoDecs use a fixed Gamma and limit the color ranges by default. This requires to manually change these options, which I guess you cannot do on the device. In Avid systems there's even a project setting for this...

           

          Mylenium

          • 2. Re: ProRes VS DNxHD
            Rick Gerard Adobe Community Professional & MVP

            Throughout the production pipeline you want to have as flat an image as you can get. This gives you the most room for correction. Buindling up contrast can lead to loss if detail in both the whites and blacks. The real test of a codec isn't what the unprocessed image looks like on the screen, it's how it works through the producion pipeline.

             

            I just ran some tests with the new GoPro H3 Black and found that the MP4 files from the camera contained a significant amount of data in the over brights that was completely lost in a 32 bit conversion to cineform, proRez444, Black Magic 10 bit. A good test is to drop an adjustment layer on top of your footage with exposure and with Effects>Utilities>CC Overbrights applied to test for detail in the highlights.

             

            goProTest.jpg

            When I ran the same test with the file transcoded to several other codecs I found that the overbrights were clipped. No values above 1 were displayed and reducing the exposure by .01 removed all of the overbright clipping display.

             

            This is the kind of codec you're looking for. Worry about gamma and timing later. Get the most info you can in your recordings so you can manipulate the image later. This clip is nearly perfect with a great deal of detail in the highlights and open shadows.

             

            Don't misconstrue this test to show that ProRez and other 10 bit codecs don't record overbrights. They work just fine. It is just that transcoding GoPro's particular version of the MP4 codec clips this the whites even when using GoPro's CineForm Studio to do the transcoding. This is one of the rare cases where I would not reccomment transcoding a consumer camera's native codec before color grading.

             

            Here are 4 frames of this clip that you can test...

            • 3. Re: ProRes VS DNxHD
              lasvideo Level 4

              I have left Prores and DNxHD behind as mastering / intermediary codecs. Instead I use MXF OP1a AVC-I00. Since it doesn't use QuickTime as wrap (or need qt32 server which creates a bottleneck ), OP1 render / exports about 2X as fast.

               

              When I was using DNxHD as the export choice out of AE, I found when configuring it in the Render Queue it was best to select RGB. That way I avoid a little gamma boost in the file when compared to the original.

               

              Screen Shot 2012-11-13 at 6.54.16 AM.png

              • 4. Re: ProRes VS DNxHD
                Pierre Devereux Level 2

                Hi Rick,

                 

                Thank you for the reply and the examples. I will take my equipment outside and do some daylight shots and run the same tests that you did.

                 

                I do have an unrelated question though - I am very interrested in purchasing a GoPro 3 Black edition - we actually sold our GoPro Hero2 just the other day. The footage you receive from the GoPro 3 - does it have the same built in problems that Hero2 had? - The terrible Barrel Lens that caused such a distroted "Fish-Eye" view, and the jello-effect caused when the camera vibrates?

                 

                I actually want to get one of those Quad-Copters and take the GoPro up for a few aerial shots for the project, and the GoPro 3 black edition has such a stunning demo reel, but when I made some enquiries about them, to the sales department, I was told that the marketing department made some "Artistic" comments on how the camera actually works, and the images in the demo are highly edited.

                 

                I dont see that as a very conforting statement, and I now wonder how much of the demo is real, and how much has been manipulated in post.

                 

                Thank you

                 

                Pierre

                • 5. Re: ProRes VS DNxHD
                  Pierre Devereux Level 2

                  Hi Lasvideo,

                   

                  I actually see that my final product has been requested to be sent in MXF OP1a file format. They state it must be encoded with a nominal bitrate of 100Mbit/s using the "AVC Intra" codec. I assume that is what you are taling about? They also state it must use the High 4:2:2 Intra Profile@level 4.1 (but I am a little unsure what that means still)

                   

                  Since I have to capture either ProRes or DNxHD with the capturing device we use (Atomos Samurai) - I suppose it really does not matter which I choose. I will run the tests that Rick did above to see if one comes out over the other, then stick to that. I generally render out TIFF sequences for all the projects, and once I have done all compositing and added effects, I take it to Premier to edit. From there I send it to Media Encoder and select MXF OP1a and the rest.

                   

                  To be honest, I am still refining the actual workflow from start to finish, as we have yet to complete our first episode! - One day at a time right?

                   

                  Our projects contain Green screen work for almost every shot, then added effects for character movements, lighting, sets and models etc. So it seems to me that it is quite work intensive, and I will need to go through the process a few times to thrash out the most time and energy effective workflow.

                   

                  Thank you for the advice, and I hope to give feedback soon on what I have found.

                   

                  Pierre

                  • 6. Re: ProRes VS DNxHD
                    Fuzzy Barsik Level 4

                    AVC-Intra is a nice codec for delivery. However, I wouldn't use it for an intermediate if quality matters. Here is some blind test 'AVC-Intra 100 vs H.264 100 Mbps':

                     

                     

                    In regard to DNxHD, as mentioned, it behaves nicely when it is set to 'RGB Levels' for 4:2:2 10 bit, whereas to '709' for 4:4:4 10 bit (which is not supported by Atomos Samurai, according to the spec). Otherwise it may not only shift gamma, but also convert some overbrights into black artifacts. Therefore, check if Samurai allows to choose DNxHD settings.

                    • 7. Re: ProRes VS DNxHD
                      lasvideo Level 4

                      Sounds good Pierre. I look forward to hearing your results!

                      • 8. Re: ProRes VS DNxHD
                        Rick Gerard Adobe Community Professional & MVP

                        The GoPro HD3 black is much better in all regards than the previous models. Jello cam is greatly reduced, exposure latitude and exposure control is much better. The lens is still wide but you can crop the image in the menu giving you less barrel distortion. Any lens that wide that isn't $10,000 is going to have barrel distortion. I couldn't be happier with the thing. I have shot tests up to 4K. The 2K stuff is simply amazing. The 120 FPS 720p is also amazing. Shooting HD at 60FPS greatly reduces rolling shutter and I can't see it at all at 48 or higher fps.

                         

                        I'm doing some under water tests next weekend. I'm also flying with the camera on Friday if all goes well. The resulting footage is amazingly sharp and really easy to color grade.

                         

                        So much for the GoPro 3 report. I'll probably blog about it soon.