Cutting AVCHD on FCP? Stop using ProRes!

[Robert Lane]

EDIT/UPDATE:

To those of who you may have seen the original version of this post, where I had announced that using ClipWrap to encode AVCHD clips into either DVCPRO-HD or Avid DNxHD would render 12x times faster than ProRes... turns out that's not totally true, and it's my fault for getting too excited and not finishing the scientific method of verification first.

What I had missed in my original testing (after finding what I thought was the golden-goose of time-savings) was that by default ProRes standard and HQ has "Render in high precision YUV" selected, whereas DVCPRO-HD does not.

If you either select "render in high-precision YUV" for DVCPRO-HD or, de-select it in ProRes then you get to near parity in render times. DVCPRO-HD and Avid DNxHD are still faster than ProRes, but not by a large margin - certainly not the *huge* margin I'd seen before.

But, I did learn yet something of note: Using High Precision YUY renders for output to the web is a waste of time. You'll only notice it in BR encodes or if you're doing a telecine process. Otherwise RGB renders is just fine.

[Craig Parkes]

Makes sense as you'll be outputting to 8 bit rgb in the end, although for grading/effects purposes you may introduce additional/more severe banding (that may then create additional compression artifacts) by not working in high precision YUV.

But I'd agree that generally time savings for web only content are going to be more beneficial than high quality mastering.

[Robert Lane]

Once again I've gone back to new camera masters just shot for a real gig and decided to put ProRes (non HQ) timeline up against my old-time favorite DVCPRO-HD, this time making *sure* they both have the 8-bit RGB render set in the sequence settings.

Turns out DVCPRO-HD (DV100) is still noticeably faster than ProRes timeline at rendering. What took the ProRes timeline 60-seconds to render took only 35 seconds in DVCPRO-HD. Both timelines were set to 720p30 and both using the exact same transcoded AVCHD clip. (ClipWrap was used to transcode 1080/60p from the Panny TM700.)

So while my first test was bungled by my mistake in not making parity with all settings, the end result is actually the same: DVCPRO-HD is faster and with zero definable clip degradation between ProRes or DV100, both visually and on the scope, most likely due to the fact that both timelines are 4:2:2 - even though the camera masters are 4:2:0 AVCHD.

So my original edict stays the same: Want to same time in editing without sacrificing image quality? Use ClipWrap to transcode your AVCHD into DVCPRO-HD and stop using ProRes.

A full review is coming of ClipWrap - probably by this weekend. That is *after* I post the optimal MPEG2 downconversions for DVD authoring that I've been promising everybody.

[Andy Mees]

Is the raster the same for both your DVCPRO HD and ProRes 720p60 clips in these tests Robert? And if possible, do you mind sharing the specs of the various hardware and software (OS, FCP version etc) that you've tested this with? Similarly, are the margins in the results the same regardless of the testing hardware and software?

Thanks for the info
Andy

[Nigel Barker]

Originally Posted by Robert Lane

But, I did learn yet something of note: Using High Precision YUY renders for output to the web is a waste of time. You'll only notice it in BR encodes or if you're doing a telecine process. Otherwise RGB renders is just fine.

I have always used High Precision YUV with ProRes as that is the default but after reading your posts I did a experiment yesterday with a small sequence that I was working on. I switched to 8-bit RGB & rendering the full sequence took just one third of the time without any noticeable change in output quality even after being compressed to 8Mbps H.264 MPEG4.

[Robert Lane]

That's why I post this stuff, Nigel, so people like you can save time and actually get work done!

Andy,

I never post minutia that like that anymore for two reasons: 1. Since all this testing/posting I do is all on my spare time and I'm not sponsored by advertisers I limit the amount of data I post to the most common denominators and "must-have" info. 2. It's a sure bet that almost nobody will have my exact system setup, so comparing systems A/B/C to each other is pointless and again, too much time for an au-gratis venture.

However if there were some hardware-specific item that would make-or-break someone else's results or usability, such as "this can only be done with "X" configuration..." then I'd make that note.

Testing results like this are all about the fact that there is a benefit to all, exactly how much it would benefit the individual is up to *them* to discover on their own system - just as Nigel did!

[William Hohauser]

Is the render time difference due to the raster size being smaller in DVCProHD?

For web projects this shouldn't be an issue. I am working on a web project right now where I am downsizing a couple of 1920 x 1080 cameras to 1440 x 720 since that's what the main footage was shot in.