best downsampling workflow from prores422 to SD

Hi all,
So I run a little TV station (PEG), and my ultimate target is always SD anamorphic, which I then letterbox for broadcast.
But I just got these sweet Canon xf300s, and FCP makes prores422 from the .mxf files they put out.
I have been taking these prores422 files into compressor and transcoding/downsampling to DV NTSC anamorphic (since this is my default sequence setting in FCP).
This is problematic because the quality is mushy (could be just my perception of DV after looking at HD originals), but most of all the processing time is KILLING me! 1 hour of media is taking ~8 hours to transcode! There must be a better way.
Would it help to start with prores light? Is there a better SD 16:9 format/codec I could be using as a target format for editing in this workflow? I still use a fair amount of DV NTSC capture media, sometimes in the same sequence with downsampled clips from this workflow. Is there a smarter choice I could make with regard to my sequence setting?
Thanks
Eric D [adhocproduction]

> I have been taking these prores422 files into compressor and transcoding/downsampling
> to DV NTSC anamorphic

Why are you editing in DV? It is the worst of all the SD formats. If you're going to broadcast, you should be using Uncompressed 8-bit or 10-bit SD, or ProRes SD.

Conversely, why edit in SD at all? You're having to convert all your footage, rather than just what you used in the edit. Why not edit in ProRes 422 HD, then convert only the final edit to SD? Or just output the HD timeline via a capture card onto SD tape?

How long are your shows? If you're working with just one hour of footage, I'm assuming the shows are just 30-to-60-second spots. If you're mixing in DV footage, then you can also "offline" at HD quality, throw in the DV clips just as reference, and then use Media Manager on the completed edit to make an SD Sequence. Or just use the ProRes HD clips in an SD anamorphic 16:9 timeline -- as long as the frame rates match. The render time will probably be less than having to convert everything.


www.derekmok.com

Processing will very much depend on your Mac(s) you could easily set up a Qmaster network with Compressor to share the transcoding. What Mac/FCP/RAID setup are you using?

Is there a reason why you don't cut the masters in HD and then do the SD master afterwards?

Certainly I would always advocate this as you will then have an HD master for future (showreels, broadcast or library footage) and not just the "mushy quality" you will certainly get from taking a 1920x1080 picture down to 720x480!

It would also mean you only need convert the final sequence length which should mean far less transcode time.

Obviously the inclusion of DV NTSC will mean a bit of upscaling and the render times will depend on how much of your sequence is SD and how much HD.

I'd also plump for a ProRes 422 SD or Uncompressed SD master rather than DV wherever possible to reduce the DV artefacts (which will certainly appear in the Reds) when you downconvert your HD footage.

If you must pre-transcode then try going to ProRes 422 SD.

Another option:

Simply cut all your footage (HD and SD mixed) on a ProRes SD timeline and render the final output.

Although I've sometimes had weird field-order issues this route using PAL ProRes SD and DV PAL footage - can't vouch for NTSC.



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

Ah D you just got in there before me!



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

Excellent info fellows! Just what I was hoping for.

>What Mac/FCP/RAID setup are you using?
Well, I am working on an iMac with a fw800 external set as my scratch. I work for a small-town city government in Iowa, so I actually brought the first (and still only) Mac onto the premises. I send my exports off to a NAS array that is raid5, but that is beside the point here.

I am editing in DV mostly because I don't know better - and my "one-man-band" situation hasn't left me much breathing space to experiment with new workflows - I have been basically clinging to something that works. Now I just need to find something that works BETTER.

>Why not edit in ProRes 422 HD, then convert only the final edit to SD?
I had assumed that SD would be less intensive on my machine, which is often also chewing on multiple exports/transcodes while I am editing. The advice you give (both derek and ben) to edit in HD so that I only downsample a fraction of the original footage makes sense on some of my programs, but not all. For example the transcode chewing right now is for a talk-show where we are basically taping a live radio show; 85%-90% of the original media will appear in the final export. I also add lower 3rds and info-graphic bumpers to the final version. Will the "downsample to SD on export" workflow cause any problems there?

Thanks so much for the advice, I am going to try (next time) doing a prores SD downsample into a prores SD sequence. Sounds like even if I don't save time on the downsample I will at least enjoy better image quality.

For other, more "doc style" programs, where I mix SD and HD source material, I will try the suggestion you both indicated - mix formats in a proresSD timeline, then render for output. One question, though, maybe this is a big one: I often expo using QT conversion, though sometimes I send to Compressor. Can either of you shed light on pro/con issues I should be aware of when making that choice?

You guys Rock,
Eric D

> For example the transcode chewing right now is for a talk-show where we are basically
> taping a live radio show; 85%-90% of the original media will appear in the final export.

That means you're not doing much editing at all. That means you don't need much computer horsepower. If you had two camera angles, then you're talking about more like 50 per cent of the footage being used. Either case, all the more reason not to waste time transcoding the raw media to SD.

> I also add lower 3rds and info-graphic bumpers to the final version. Will the "downsample
> to SD on export" workflow cause any problems there?

Yes and no. Which editing format you pick will determine how the graphics should be spec-ed, whether they should be interlaced, etc.

Your best approach is still a capture card; bypass software conversion of the media altogether, and output an HD timeline directly to SD tape.

> I often expo using QT conversion

No, export as a full-quality movie file (Export - QuickTime Movie) first. That gives you a flattened, top-quality file with which to do tape outputs, or as a basis for other conversions. FCP is one of the most inefficient tools to use for compression.


www.derekmok.com

Quote

Your best approach is still a capture card; bypass software conversion of the media altogether, and output an HD timeline directly to SD tape.

Bit of a problem when using an iMac as you have no Card slot and external FireWire devices that can do Hardware SD downconversion such as the AJA IO HD will not play nice with FireWire Media Storage on the Same FireWire Bus (eg connected to the iMac at the same time).


So you'll have to go the software route. That is unless you have one of the new Thunderbolt iMacs? You could get an AJA IO HD via FireWire and a Thunderbolt RAID/HDD for your media drive.

If you have an iMac with USB 2 you could try a fast USB2 HDD plus an AJA IO HD but I would test a prospective USB2 case thoroughly before hiring or buy an IO HD as USB2 is not what I would consider a reliable video media connection. I know people who DO use it but I'm not going to stick my head on the line and say you should!



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

> Bit of a problem when using an iMac

My bad. Missed that part. But this is why post-production stations don't use iMacs.

He could try an I/O box while moving a self-contained full-quality movie file off the external drive and thereby taking it out of the equation. But that's contingent upon his internal drive being fast enough and consistent to handle whatever SD format he chooses. I wouldn't bet on USB either.


www.derekmok.com

> Bit of a problem when using an iMac...
Yeah, my iMac is a limited tool, but it is what I have to work with, for now. I am digging the advice to go to using "expo QT movie" to get a flattened ProRes SD .mov, and then doing whatever end-target conversions/transcodes using that master. Derek - Yeah you right - FCP (QT conversion) seems like a silly choice for compressing output files, especially since it ties up the work-station - I think I will quit it...

>Thunderbolt iMac...
Noo, no thunderbolt, this machine is ca. November 2009. And I too have learned to NOT depend on USB 2.0 performing as promised.

Wow, you guys have really helped alot - helped me to rethink my workflow habits like QT Conversion and DV NTSC anamorphic. I will update this thread in a few days when I get into the thick of it with the changes.

Thanks again
Eric D

No problem Eric - hope it smooths out the flow or at least increase the quality of the output.



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

Hi Derek and Ben,
Just an update on the issue you guys were advising me about a few weeks ago - here's what I am doing, you guys tell me what you think.
OK, so in order to avoid the time-killer of transcoding all clips to SD up front, I am now using a customized 720x480/anamorphic pixel/ProRes codec sequence. This allows me to pull in and edit both my DV NTSC anamorphic AND my 1080i ProRes422 clips in the same sequence. The HD stuff is "green-lined" but at least I can get to work and let it render on idle time. The rendering here seems to take about 60% less time than transcoding the same amount of footage to DV, and that rendering is much more convenient b/c it doesn't hold up progress. This is HUGE progress, and amounts to a big fat "PROBLEM SOLVED" for you guys, with respect to my original raison d'etre for starting this thread. Now, I assume that there is some increase in image quality, especially in my reds, since I will be able to move images through this workflow without them ever losing color data by being transcoded to DV, but I have not had an opportunity to really evaluate that - I kinda see it, but I really need to view it on a proper monitor to be sure.
Anyway, here is a new wrinkle you might find interesting: I took a finished piece that was on a ProRes SD anamorphic sequence and sent it to compressor twice, one job set up to output ProRes SD anamorphic (setup identical to the sequence), and one job set up to output DV NTSC anamorphic. Batch Monitor shows that the 1:40 (hr:min) program took 5:59 to output to ProRes, but only 1:24 to output to DV. Maybe I am skewing the result somehow with my not-that-sharp skills at setting up compressor profiles, but that is the result I got on this run.
Once again thanks for all the help, and especially just being here - it is encouraging to know there is a place to ask these kinds of questions, and find answers from knowledgeable people.
Rock on
Eric D

> I took a finished piece that was on a ProRes SD anamorphic sequence and
> sent it to compressor twice, one job set up to output ProRes SD anamorphic
> (setup identical to the sequence)

If you're exporting at the same quality and codec as the Sequence, why use Compressor at all? Just use Export - QuickTime Movie, Current Settings.


www.derekmok.com

Good question, Mr. Mok. In a perfect world I would do just that; I do now see the elegance of the simple "current settings" expo function from FCP.
I am not using it in this case, or in most cases, because I have only one FCP workstation, and several projects live with deadlines ticking: that is just to say that most of the time I can't afford to have FCP tied up and be unable to continue working on other projects. Therefore I send the project to compressor to expo in the background while I keep working.