Alexa: ProRes vs. DNxHD
I'm advising a production that is considering using the Alexa for an upcoming TV show. We have experience shooting to ProRes 444 and are intrigued by the DNxHD 175x option (we'll be cutting on Avid); however, our director of photography is concerned about image quality. I've seen the tech specs and I know ProRes 444 is a beefier codec from a number of different measurements but I'm not sure how that translates to the real world. I'm particularly interested in how the DNx image holds up when zoomed in on in post as compared to the ProRes media.
For example, if the intended final delivery is for HD television, you'd feel comfortable blowing up the ProRes to xxx% of the original whereas you'd only want to blow up the DNx to yyy%.
I've also heard that the Alexa will be shooting to DNxHD 444 but I haven't heard if they actually are capable of this yet. Anyone have any experience with that? Thanks, guys!
I don't think the DNxHD444 has been given to camera manufactures yet, or it has not yet been implemented. Maybe NAB will bring news on that front. Also keep in mind that not all of Arri Alexa's come natively with ProRes - DNxHD is an additional cost (may have changed but was $5K option when introduced).
As far as quality differences, it is always a more difficult and subjective discussion as it is base on what is being shot, etc. I don't know of any codec shoot-outs between ProRes and DNxHD yet that have been done. But assuming both are being shot at 1920x1080, it's going to be very close as far as percentage of zoom into the image.
The differences will be 444 RGB versus 422 YCbCR from a colorist standpoint and to some extent green screens and such. ProRes is 12bit, whereas DNxHD is 10 bit. That is a discussion to have with your colorist.
Which brings the conversation to what is the post workflow for color correction and mastering? This is where some advantages might happen depending several factors. The main advantage is to avoid an offline/online step and conform is a consolidate at best. But if the DNxHD was shot with LogC, you may need to do a dailies pass anyway if editorial wants image back to Rec.709.
Both processes will require AMA link - for DNxHD you want to do so to get all the metadata, and for ProRes you have to. The ProRes has an extra step as the QuickTime AMA does not extract any metadata out of the QuickTime other than timecode. On each card, there will be an ALE created by the camera that you can merge -
ProRes is now a native codec within Media Composer - on OSX you can also render to that. You would need to work in a RGB project at the 444 data rates in both cases to preserve the high resolution throughout. Now it is a matter of do you have the storage and speed requirements to satisfy the type of editing this will entail? For example, is it multicam, lots of layers, etc. Performance will be affected in some manner unless you plan for that.
From an "archive" point of view, I would tend to go 444RGB over a Rec.709 codec if given the choice - regardless of whether it is ProRes or DNxHD. Also, DNxHD 444 takes up more storage space on the recording card or editing storage compared to ProRes, so keep that in mind as well as it affects productions, number of cards, transfer time, etc.