Hello all,

We’ve been using Premiere Pro CS5.5 for about a few months or so after ditching FCP7. We shoot primarily H.264 footage right now and will be getting a C300 in in a few weeks. Even though PPro will edit with H.264 we decided to go with our old workflow of converting to ProRes422 for edit (after trying to edit in H264 and getting some slowdowns and stuttering). We have also experienced some major slowdown (compared to FCP7) with rendering while editing in ProRes422 – anything more than one layer of video.

My question for everyone is: For a large project, do you convert all footage to one common denominator or edit everything in its native format?

If you do convert, what codec do you use?

It has been a bit maddening to have to render after every little adjustment in the edit or small tweak on a filter and scale, etc. (Final Cut would need rendering as well - but at least it would play it back at a lower resolution).

If you do something else entirely, please post any workflow ideas that may work better.

Any thoughts would be helpful and appreciated.

Thanks,

~ Marc


MacPro 2.66 Quad-Core Intel Xenon *** 8 Core coming soon!
16GB of RAM
OSX 10.6.8
NVIDIA GeForce GT 120 ***Quadro 4000 coming soon!
Caldigit HD Element 4TB RAID (via mini SAS)

Native format editing with mercury playback enabled.
I used to convert everything to ProRes422(LT), but not anymore.

I usually right click (or control click) and lower the quality to as low as possible under pause and playback, but will check it at full quality when the edit is done.

For color grading I dynamically link to an AE project with 32bpc and use Colorista II and Magic Bullet Looks 1.4

My typical shooting format is R3D, but I recently cut a feature that was Canon h264 flavor.

Here's a good guide for doing the mercury hack, which people might want to keep for reference. It also will show you if your card is not recommended for the hack:
http://www.studio1productions.com/Articles/PremiereCS5.htm


But even with Mercury enabled, there are some things to take into consideration.
To quote the article:
Number 4 - Decoding
The decoding of video footage is still handled by the CPU and not the GPU. So if you are using heavily compressed video, such as AVCHD or h.264, you will still need a fast processor, since the video card won’t help with the decoding of the video. Also, keep in mind, that if you use a non-Mercury Playback Engine enabled plug-in or effect, the GPU on the video card won’t help you with rendering that effect.

Yes, we are upgrading to the Quadro 4000 - actually we have it in house already. We are just waiting for the 8 core tower to show up.

Unfortunately, the only cards that work in a Mac and are supported by PPro for hardware acceleration are the:
GeForce GTX 285 (discontinued)
Quadro FX 4800
Quadro FX 4000

So we went with the Quadro FX 4000. We'll so how much of a difference it makes.

So considering that the most commonly used cameras these days are acquiring video in pretty heavily compressed codecs - the question is:
Just because PPro can edit natively in those formats, is it wise to do it that way - In terms of a post workflow (color correction, SFX, etc.)? Is there a 'best' codec to edit in that will take the load off of the CPU.

That's really what I'm trying to figure out while we are doing an overhaul of gear and trying to figure out a new workflow for PPro.

Which is the nobudget autodusk lustre smoothly operating graphic card ?

I have been trying to establish exactly what happens and how and the word is that there is absolutely nothing gained converting to ProRes*. PP converts to its own proxy that is 4:4:4 with a 32 bit float where supported (all NLE’s have to convert to baseband 4:4:4 in YUV or RGB depending their code processing abilities - I am told).

* unless you know differently!

I think there is something gained converting to ProRes. I haven't done testing myself but I seem to remember Mr. Bloom doing some way back when. I think there was some quality gained if I remember correctly. Maybe if he sees this he will chime in...

I've also been looking back at software like 5DtoRGB - http://rarevision.com/5dtorgb/ - which I totally forgot about. Great quality but a bit slow.

Let's try and nail this once and for all. Is the PP proxy good enough to allow latitude when grading or is there something to be gained by converting to ProRes? What we need are FACTS. How do we get them?

I've nearly the same set up as what you'll have. I shoot on Canon 7D's and also Sony Z7's and I don't convert anything. Plays back great even at full resolution, color correction, effects etc. You can of course drop the playback screen res if need be.

There's always trade offs, no matter which way you go. H.264's are quite CPU intensive to playback due to the type of compression involved. To view a single frame, the codec often needs to look at at multiple frames before and after the current frame in order to correctly 'build' the frame you want to look at. Something like ProRes on the other hand, whilst taking up a lot more disk space, is technically an easier codec to work with as it requires less CPU power to decode. I have found (working on a MacPro without the supported graphics card for Mercury Playback acceleration) editing with H.264s to be 'unresponsive', especially when trying to play backwards at any given point. There's nothing wrong with editing straight from the H.264s (or any other camera format) in terms of image quality, at this point it is more a question of performance and disk space.

In terms of quality there's certainly no way to 'improve' the quality of footage by transcoding it to a 'better' format. Think of it like a pipe with water flowing through it. The narrowest part of the pipe will determine the rate at which the water flows. If you widen the pipe at a later point, you won't increase the amount of water flowing through, it will still be limited by the narrowest point. You can take H.264 and convert it to uncompressed DPX's if you really want, but that will just be a huge file, with no more actual data than what was in the original files.

Someone previously asked "Is there a 'best' format for editing?". That's a tricky question, as it depends on what your requirements are. You might need something that takes a load off your CPU, or maybe something that takes a load of your hard drives. You might want something to give you the highest image fidelity through your pipeline, or you may be happy to work with low-res proxies while you edit. For me personally, with Premiere, I edit natively in the acquisition format wherever I can, to save on time spent transcoding, and the extra hard drive space that goes hand in hand with having an additional copy of the footage. As soon as I am exporting or rendering out something, be it motion graphics or grades from After Effects, or a final copy from Premiere, I make sure I'm exporting a file format that is of high enough quality. For me, that's ProRes 422 LT, which has more than enough data-rate for DSLR footage. Other's might prefer straight ProRes 422, or Avid DNxHD, or other great codecs like those from Cineform.

Once I have this final file, I use this to make any deliverable copies (DVD, mp4 etc). This way I have a high quality file which I can archive, backup and use to make any other copies I need. Because the ProRes file is over a greater data rate than my acquisition format, I know I'm not trimming off any important data. Once I have this file I'm happy to start throwing away data to compress it for DVD or the Web.
I've seen editors grade a project in after effects in 32bpc colour, using Magic Bullet Looks or Colorista, and then render out a low bit-rate H.264 to take back into FCP, so that it 'matches the source footage'. This is a horrible thing to do. They just took their high fidelity image and squished it down, throwing the majority of their data that they just created and shoved it all in to a compressed codec. H.264 is not an editing codec; it is a viewing codec. It is designed for final delivery, it is not designed to have fine colour adjustments made to it, or any of the other subtle things we do with our footage. Once you've made changes to your footage, and you are rendering or exporting, this is when you need to be concerned about file format.

If you're planning on round-tripping your project through a lot of applications other than Premiere (perhaps After Effects, or DaVinci Resolve for grading) then I'd definitely suggest transcoding your footage to either ProRes or DNxHD, as these are more widely supported, and often have performance gains with other applications compared to native acquisition codecs.

Cheers
Jacob.

Thanks for your input Jacob.

Great post, Jacob, but I have a question. I was wondering about speeding up my CS5 playback by taking my MTS footage and converting it to a format that requires less de-compression to play back. Is there a format that accomplishes this?

Or would using a RAID system improve my playback more effectively?