Apple ProRes - which version based on these edit requirements?

Hi, first and foremost I do understand that the whole 'ProRes 422 vs ProRes 422(HQ)' debate has been covered many times in the past, so apologies in advance for digging this up once again! I've been reading up on this over the past week or so and I can't find any definitive answers specific to my issue, so I thought it was best to post it up on here.

I'm due to commence editing a short drama film soon which upon completion will feature on the festival circuit throughout next year, with a potential TV broadcast in the works too. We shot the film predominantly on a Canon5DmkIII and an FS700. I will be editing the project in FCP7.

My question here is that I'm a little unsure as to which codec to transcode the rushes to for the edit...all of the footage was shot 'flat' which means I'll need to do some extensive colour grading in post, as well as additional sharpening and de-noising (Neat). Normally for my day-to-day editing which consists of mainly straight to web content I would transcode everything to ProRes422 or 422LT, which is obviously more than good enough when delivering the work online. Once again I'm considering converting everything to the standard ProRes422 codec, but due to the amount of post work that needs performing and the fact that the film will be shown on potentially a cinema screen and on TV I'm wondering if converting to 422HQ is worth considering here based on the edit requirements and final output?

I mention the HQ codec in relation to its data/bit rate; meaning that as HQ has a higher data rate than the standard ProRes codec there's more 'head room' for retaining quality & detail after rendering the footage out and then exporting? Apologies if I'm wrong in thinking that but that's pretty much my enquiry in a nutshell there. I understand that converting to HQ initially will not give you a better quality file than the standard 422 codec (overkill in most cases), I'm just wondering if HQ (or any other codec for that matter) will retain it's quality better after some potentially heavy grading & filtering vs. codecs such as ProRes422. Storage space and render times are not a problem here, but I would only convert to a codec such as HQ if I know it's going to be worth it in the long run.

Any help/advice on this would be much appreciated. Thank you.

Alex

I am one if the ones that says go for ProRes standard in most cases, but in your case I agree HQ is the way to go. These are the exact reasons to use HQ - storage not an issue, doing extensive grading, and possible delivery to high end broadcast. Many of them in fact require ProRes HQ delivery these days.

yeah, i'll second HQ in these circumstances.

nick

Alex, the Apple ProRes White Paper is a good place to start learning about that family of codecs. This graph on page 14 is relevant.


For a particular image -- Apple doesn't say which -- the first ProRes HQ copy of that image has a small pixel-by-pixel luma disagreement with it shown by its PSNR of 53.5. For perfect agreement PSNR is infinite, but 53.5 is quite good. 53.5 implies that pixel-by-pixel, the (rms) average disagreement between copy and original is ±2.2 counts of the full 0-1023 10-bit scale. Continuing with the graph, after recompression to ProRes HQ the average disagreement with the original is ±2.7 counts. After another recompression to ProRes HQ the average disagreement with the original is ±2.8 counts. For comparison, when using plain ProRes the first generation disagrees with the original by ±4.0 counts, which grows to ±5.0 counts and then to ±5.4 counts.

Hopefully, converting the dB scale to "±counts" (for the 10-bit scale) gives video people a better feel for this PSNR measure.

70 dB means ±0.32 rms
60 dB means ±1 rms
50 dB means ±3.2 rms
40 dB means ±10 rms
30 dB means ±32 rms

The graph shows how for copying, recopying, recopying, etc., image decay ceases after the third generation. It is a good compression codec that can do that, but it does not imply free-sailing through many generations of filterings. Filterings generally produce new images. The new images can even be sharper than before. Consider that ProRes HQ can hit with ±2.2 counts of new disagreement at each newly filtered generation. Consider that plain ProRes can hit with ±4.0 counts at each generation.

The newest ProRes XQ fits into the gap between ProRes HQ quality and Uncompressed quality. Unfortunately, the White Paper includes no PSNR data for ProRes XQ.

The matter of 4:2:2 vs 4:4:4 codec needs to be examined too, even when the original is 4:2:0. Topic for another strand.

Dennis Couzin
Berlin, Germany