Optimizing Media-when to do

[Tony Silanskas]

Converting DSLR footage to ProRes the last few years, I’ve yet to see a noticeable drop in quality. ProRes is a very solid codec and since FCP X is built around it, it will give you the smoothest editing situation. I won’t go into details here as this has been discussed to death on the forums but h.264 was not made to be an edit-friendly codec. It’s highly compressed and made for delivery and therefore eats a lot of CPU/GPU resources to edit on the fly. But FCP X has made huge strides in using every ounce of your computer’s power to help in that area and will only get better with Lion.

Also, from my understanding, FCP X now renders everything at 32-bit floating point (instead of 8 or 10 like before) so that will help keep quality up. Even if you edit h.264 natively, it is still being converted to ProRes in the background anytime you add effects. I can see that you are worried about converting to ProRes first and then converting to ProRes again when adding effects but like I said earlier, ProRes is a solid codec and you now have 32-bit floating point rendering so you’ll be fine. For me at least, I’d rather gain smoothness in editing until I upgrade to the latest and greatest MAC and Lion that may handle h.264 as well as native ProRes.

tony

http://www.HungryCliff.com

[Jim Bachalo]

Thanks Tony.
I do have the latest and greatest iMac, so will probably do my own test to see if I notice any performance issues without optimizing media.
Of course, this means I still must purchase FCPX…at the moment not sure if it will be useful for anything other than a prototyping tool. Everyone has their own list of what’s missing, but for me its RGB curve-LUT support when color correcting and Bezier key framing control when time remapping.

Local is the new global

[Cyrus Dowlatshahi]

In his tutorial, Larry Jordan suggests that converting to ProRes improves image quality:

“The file sizes are bigger, but the color quality and image quality is greater. Also the performance of Final Cut is better.”

I’m doing some side-by-side testing using the J-K-L keys to fast-forward and rewind through native footage (in this case AVCHD) and also footage transcoded to ProRes. My computer (an 8-core Mac Pro) handles the ProRes noticeably easier, but compare the file sizes of a 38-second clips in AVCHD (80.8MB() versus ProRes (525.5MB).

I can deal with some choppiness as I edit, but I don’t want to sacrifice image quality. Image-quality-wise, will a final sequence edited natively look worse than if it was edited in ProRes?

[Craig Seeman]

[Eben Abbaan] “”The file sizes are bigger, but the color quality and image quality is greater. Also the performance of Final Cut is better.””

Does he give any more details on that? It’s possible that some interpolation “smooths out” some of the codec issues. Certainly the clip should hold up a bit better when adding any FX.

[Cyrus Dowlatshahi]

I don’t know the insides of codecs that well.

I’m pretty sure than any portion of a timeline to which an effect is applied is converted to ProRes (per FCP X’s render settings).

[Nate Weaver]

I disagree pretty strongly when people talk about transcoding to Pro Res is going to increase your image quality.

You can’t make a better image (i.e. less compression artifacts, get the benefits of a 10bit file from an 8 bit codec). Just like when you upscale SD to 1080p. More detailed image data doesn’t just magically appear.

You’d be putting 6lbs of picture in a box that holds 10lbs. But you don’t get a free image quality lunch. The damage is done when that uncompressed baseband video inside the Canon is compressed. It doesn’t come back. Not a tiny bit, not a lot.

Nate Weaver
Director/D.P., Los Angeles
https://www.nateweaver.net

[Craig Seeman]

It may not be that simple. The “damage” the interpolation does (the estimating of information not there) might actually improve things.

I’m trying to think of an analogy. Although an imperfect analogy, it may be like how adding noise gets rid of banding. It may be like the various filters that improve keying in 4:1:1 footage. Sometimes making something a bit “mushy” actually may subjectively improve something.

The comments about the improvement in quality made above provided no detail, not even a subjective analysis. But what if the viewer felt that aliasing and/or moire didn’t look as bad as a result of the uprez. They may have softened a bit.

It may well be that the degradation of an artifact may seem to be an improvement subjectively although objective it’s creating more damage.

Larry Jordan is by no means a novice so I don’t doubt he’s perceiving something subjectively . . . assuming the quote is accurate.

[Cyrus Dowlatshahi]

FYI, The quote is accurate. Here’s what he just wrote me:

“Will picture quality suffer [by editing native h264/AVCHD]? Not particularly. The downside is render time and limited color correction space.” -Larry

[Craig Seeman]

[Eben Abbaan] “limited color correction space”

Thanks for that.

It’s why I said
[Craig Seeman] Certainly the clip should hold up a bit better when adding any FX.

[Nate Weaver]

[Eben Abbaan] “”Will picture quality suffer [by editing native h264/AVCHD]? Not particularly. The downside is render time and limited color correction space.” -Larry”

Render time nobody can argue with. “Limited color correction space” is what does not make sense based on what I know of image processing in Resolve and FCP (or X).

The 8bit codec is decoded into a 32bit deep frame buffer, whereupon processing is performed in 32bit float. At the end of the pipeline, it is truncated back to 8 bit or 10 bit, depending on viewing method or render format.

Larry’s assumption is that the frame buffer/processing is 8bit/4:2:0 because the codec is 8bit/4:2:0. Not really the case, not in FCPX, not in Resolve. Frame buffer is gonna be 32bit float 4:4:4.

Nate Weaver
Director/D.P., Los Angeles
https://www.nateweaver.net