HD>SD downconversion Mac/FCP only
 

I've been totally confused (and alarmed) by all the threads about HD to SD downconversion recently and thought it might help if we could at least separate the discussion by platforms and editing systems. Right now anyone facing this problem would just get buried. The aim eventually should be getting a few stickies here that people could reference.

So for a start how about making this one only for Mac OS. Further I would like to get suggestions for 2 separate issues.

1. Downconverting to an SD DVD.
2. Downconverting just to SD quicktimes. We all have clients who actually might have preferred to just shoot in SD or need footage to cut into an SD show, but we may have only an EX-1 to offer them. I would like to be able to hand them downconverted DV files or say " "Just do ...... with those files I gave you."

Another set of threads for other platforms and programs could run separately.

Leonard Levy

video seems to be missing sound?

Probably doesn't matter unless you speak German.

Easy solution. Buy a copy of Cinema Craft Encoder MP.

Sorted!

CINEMA CRAFT Encoder SP2 - OMNI-CINEMA CRAFT

This video was actually just a quick test of youtubes new HD-capability, so I didn't record my voice.

Cinema Craft sounds great, but I can't spring $800 right now just to be able to downconvert.
If I was a post house it would be no big deal, but I'm just a poor cameraman.

Dominik - I haven't looked at your you tube post yet, but was wondering if it details a good method for making SD DV quicktimes as well as going straight to DVD?

- Lenny

Here's a much cheaper alternative ($39).

CRAMCompressor

Thanks Leonard for suggesting this idea! The multiple posts are confusing!
Maybe the "best" of these suggestions could be in the stickie.

Evan

The Matrox MXO (version 1 - not MXO2) is the best hardware solution I've come across for on the fly down converting. You can tell the MXO to down convert your FCP timeline and use the component output cables to a stand alone DVD recorder. If you have another FCP system, you can down convert out from the MXO straight into a SD final cut capture card, thus creating your SD QuickTimes. I do this on a daily basis for my clients.

The MXO in my opinion is the easiest, no hassle way to down covert any HD format on the fly (system requirements apply). Matrox have created an amazing product with the MXO, as this was just a bonus feature they threw in. Unbelievable!

-C

Had a couple of days playing around with Cinema craft Encoder MP. Of course one small problem I do have is that I shoot weddings at 720P50, so when I encoded in Encodr MP, everything is 50% speed because it compresses at 25 FPS.

So I exported as a quicktime coversion H.264 best 25FPS Pal 16:9 and bring that into Cmpressor and then out via the Encoder MP and went with a 10 pass VBR.

I then ran a straight compressor version 2 pass VBR and compared the two when I dropped them into DVDSP.

The Encoder MP version looked better than the straight Compressor version. However, I am not happy with having to convert to H.264 so there must be a better intermediate codec to use which is lossless and allows me to go from 50 frames a second to 25FPS.

Thoughts?

Of course. The only difference is that you choose DV instead of mpeg2 as output-format.
But you shouldn't use DV as an intermediate codec, because it uses different color-sampling compared to progressive mpeg2. So you would loose chroma-resolution. Also DV is not free of compression-artifacts.

Christopher,
That's an interesting idea. I have a copy of MXO though I've never successfully set it up for video monitoring. ( I probably just haven't worked on it enough.)
Could you use Matrox MXO to output SD to either a DV recorder ( or camera )with either Y/C inputs or component inputs thus allowing you to go from the DV tape to DV quicktimes?
How good would this quality be?

Dominik,

"But you shouldn't use DV as an intermediate codec, because it uses different color-sampling compared to progressive mpeg2. So you would loose chroma-resolution. Also DV is not free of compression-artifacts."

Would that DV look worse than footage originally shot as DV in another camera?

I'm thinking about the situation of shooting footage on an EX-1 for a client who actually would have preferred a DV camera in the first place. In this place it would be an intermediate codec.

Lenny

It will look better.

Hi Christopher, sounds good! But why not the MXO2 version?

I tried something similar by using the EX3's SDI SD output to play out (720P50) and record a 1 1/2 hour program straight to my DVD recorder> I used my Blackmagic Declink SD card so I could monitor the SDI signal/sound. Worked well, although I had to use the DV D recorder's S input (best it has) plus do bit of post tweaking to the VOB file via MpegStreamclip (losslessly I believe & quick!) to get a m2v/ac3.

Saves a lot of time, especially for a long recording. Not sure about using this method for shorter projects as although the SDI output with SD downconvert was the best signal out I believe, I'm not sure it beats other software methods.

Do you think the MXO would be better?

Evan

When using compressor you should activate the advanced Frame Controls and select the "better" or "best" resize filter. Turning Anti-alias up to around 40 or 50 reduces any jaggies.

Doing this makes a huge difference to the quality of any frame resizing such as from an EX to SD. You can set this up on any preset so you can go straight from the timeline to m2v for DVD or DV or H264 in one hit. Renders do take longer but you only have to do the one step.

It's been mentioned several times that Compressor's Anti-alias feature under Frame Controls should be used when scaling down. But it's always been my understanding that this feature is only used when scaling *up*, so I've not tried it for that purpose.

Curious. Has anyone ever a/b tested downscaling behavior while using it?

I just tested some combinations of quality-settings and anti-aliasing-settings with 10 seconds 1080p converting to 720x480p. I attached the results to this post in this order:
quality=better, anti-alasing=0 => rendering-time: 0:27
quality=better, anti-aliasing=100 => rendering-time: 4:27
quality=best, anti-aliasing=0 => rendering-time: 0:33
quality=best, anti-aliasing=100 => rendering-time: 7:37

I actually can't see any difference in quality, but a lot of difference in rendering-time.
Btw, compressors manual says that the anti-aliasing-feature is only to be used for upscaling.

Now I tested converting the result of quality=best, anti-aliasing=0 back to 1080p with some different settings:
quality=better, anti-alasing=0 => rendering-time: 0:41
quality=better, anti-aliasing=100 => rendering-time: 1:33
quality=best, anti-aliasing=0 => rendering-time: 1:20
quality=best, anti-aliasing=100 => rendering-time: 2:19

The last attachment is the unprocessed 1080p-source.
I had to decrease the jpeg-compression-quality to be able to upload the pictures to this forum, so there are a lot of jpeg-artifacts, especially in the dark sky.

I read a post in here somewhere that said: Edit in the Sequence settings that are as close as you can come to matching your final output. In other words, if you aren't going to author a Blu-ray disk or some other HD final product -- then don't edit in HD!! Edit in SD. Choose sequence settings that match your final output -- not the source footage.

Ok this makes sense I think but what are the settings for doing this.
My last shoot was 1920x 1080i pal
So what should my my easy setup be, and also sequence settings be?


Simon

I edit in HD, rendering is not necessary and add my HD sequences to the SD timeline. Then I have to render and after that I create the .m2v files.

Unless I'm confused (quite possible), the timeline determines the output but you can drop whatever you like into it (in current FCP) and it will edit at the resolution of the inserted clip. You can zoom/pan within an HDV clip (up to about 4x) in an SD timeline and still get full SD quality output since the HDV resolution allows for this.

MXO2 has the same feature. I just wanted to differentiate between the two products as their seems to be large price gap and utilization difference.

Let me just be clear that your workflow with the EX3 is different than the workflow I was mentioning, as I believe from your post that the EX3 was doing the downconversion to make the DVD dub. This wasn't what I was getting at. The MXO can do this directly from the FCP timeline in its native HD form. So you would transfer your digital EX3 files into FCP and then play out via the MXO.

As for your second question about Method - this works just as well for shorter projects in my opinion. Again, if you do a bunch of these types of "on the fly" conversions, I'd totally recommend getting a second system just to capture.

Editor 1 w/ MX0 (Player) >
Editor 2 w/ SD Capture Card (Recorder) >
SD Monitor out of Editor 2 (NTSC / PAL)

There is another side to this workflow - frame rate - lets say I shoot everything 23.98 and need to make it NTSC. The MXO also will add the necessary pulldown. I haven't tried PAL yet so I'm not sure if it will convert that.

As for quality - SDI is SDI , Analog is Analog - We could debate whether the picture would look better if you stayed digital but it would be negligible even to the best (if your codec stays the same). Ultimately though, I would stay away from S-VIDEO. Component output from the MXO will give you an improved image over that Y/C junk. I liken S-VIDEO with Composite.

Hope this helps,
-C

If software-methods can be free, perfect-quality, easy to handle, flexible and fast at the same time, then why use hardware-methods?

Interesting comparisons, Dominik.
Thank you for taking the trouble to post those.

Ever since I've been doing video, it's been generally accepted that Hardware conversion is better than Software conversion.

With this in mind, I would certainly like to try Hardware conversion, but for some obscure reason nobody seems able (or willing) to suggest a suitable Deck for converting to.

I've had plenty of suggestions for Blackmagic, Kona and Matrox to convert HD but what should I stick on the end of it to catch the SD?

It would seem the deck needs Component or SDI (In) and I assume these Decks are pretty expensive.

I only want to use the SD for DVD.
Any suggestions, anyone?

Andy, why stick anything on it? If you are playing into an Aja, or Blackmagic card, just let it save the files to your hard drive, and make your DVD from there. Why go OUT to a deck, which you'd then have to digitize back in the computer?

Maybe no one's suggesting anything because it's totally unnecessary.

That thinking is really silly.
hardware can't be better than software in terms of picture-quality, because the thing which does its image-processing-job in the hardware is also just a piece of software, but software with realtime-capability-restraints.
Software on a PC can use as much time as it wants to do its job, so the programmer can concentrate much more on picture-quality.

Exactly. Hardware is for people with time constraints and deep pockets. I invested £20k in Digital Rapids encoders a while back (for encoding 300+ hour long DVD-Rs to three streams of Real Media each), and they paid for themselves in 4 weeks. I tested the whole job out using a software codec (IIRC something like 'Moonshine' or whatever - some talented Russian developers anyway) and the software version was better than the Digital Rapids, but took forever.

Business case for both methods.

I have heard from colleagues working for Auntie Beeb that Compressor, when given annual leave, can do amazing things that a quarter of a million quid's worth of Alchemist sweats over.

We've just got to balance deadlines, budget and quality.

I'd look very carefully at Matrox hardware and progressive material - there's some interlacing mojo going on that we didn't pin down.

Everyone's talking about converting their footage to SD. What about converting a finished project with graphics, titles, etc....? Is no one interested in that? (other than me)

I would like to shoot and edit in HD. That way, when my client comes to approve the project, they will be seeing it in glorious HD on our 24-inch JVC HD monitor. I could then give them a Blu-Ray version of the project (if they have access to a blu-ray player....doubtful)

But then, I'd like a good solution to converting the entire project to SD so I could:

1) Make a DVD version of the project
2) Make a MOV version of the project to send to the local broadcasters
3) Make an WMV or Flash version of the project. (might be able to do this in HD...dunno)

I'm hoping that Compressor is going to be the solution.

The quality of a rescaler is measured by how near it can approximate the nyquist-limit without producing aliasing. That specification is independent of the type of image-source.
So yes, compressor does a great job with graphics and titles, too.

Very interesting Dominik. I'm embarrassed to admit that's a bit over my head, but interesting none the less.

I'm glad that you feel Compressor does a great job with graphics and titles too.

SOLUTION!! I just remembered that our company subscribes to the training website Lynda.com (only $25 per month....well worth it!). Anyway, they have a training video titled Compressor 3 Essential Training with Larry Jordan. The "solution" was at the end of the video is the following chapter:

CHAPTER 15: Transcoding
-Understanding Transcoding
-Converting HD to SD-maintaining a 16x9 aspect ratio
-Converting HD to SD-using a 4x3 center cut
-Converting HD to SD-using custom frame areas (really cool!)
-Creating a 16x9 letterbox from 4x3 footage (stupid)
-Converting HD to ProRes
-Converting NTSC to PAL
-Converting PAL to NTSC

I watched all of them except the NTSC to PAL versions and it looks like Compressor is the solution! He has some very specific settings that he changes to make it work, but it's all based on the Apple presets. Unfortunately, I'm at home right now so I can't do any testing to see how good the quality is. (sorry if this sounds like an ad for Lynda.com, but for $25, I figured you would want to know)

Me too - but I don't mention it because I didn't want to sound like I was introducing the Holy Hand Grenade of Antioch with nods to duck-billed platipode, oranutans and bacilli et cetera. There's lots to be aware of with down converting motion graphics et al. In fact, EX1 footage has a lot in common, being rather sharp in its natural form.

After all, there was - once upon a time - rules regarding colours, font sizes, line thicknesses, pixel/scanline-per-second speeds. Strong colours in geometric shapes in a 4:2:0 codec were a no-no. Lots of things to worry about other than the need to blur before big downscales.

But I digress.

Warning: General Rant Mode is ON.

Compressor is a very good solution, though it is a widely misunderstood solution as it is far too easy to mistake a pep-talk - "compressor, take this movie and make it look really good" with a mission - "compressor, take this movie and do everything you can possibly do, including disturbing the fundamental physics of reality and upsetting the space time continuum, to make this the best it could ever be".

Compressor is very powerful - not in a Thug kind of way, but in a Commando/SEAL kind of way. So when you ask it to nail something, it sometimes decides to plot the vector of every pixel in your three hour long opus, which is why one needs to know how to chop up a long job so it only does the Mission Impossible bits on what needs to be done, and take the bus everywhere else.

There's a great little training package from Ripple Training that opens your eyes to the mentality of Compressor. Highly recommended. Then its a case of Caning it round the track to understand when to switch the gear on.

I am highly excited by the CRAMcompressor pack previously mentioned, though aware that with 300+ presets it may require a manual to enable you to choose which one.

But Compressor is up to the job. I have Episode, Squeeze, Compressor, DV Kitchen, MegaPEG Pro, Flix Pro, and a whole lot of others, and in the last few months I am finding the world increasingly compressor-shaped. Mainly because I think I'm finally understanding it. Or grokking it, if we're now dropping Nyquist into this...

Am I going to step into this Mac discussion, and I ask this question truly out of ignorance.

On the NLE's I've used on the PC, We drop our video (of whatever source size onto the timeline. The graphics, titles, etc. are laid on in separate tracks. We then render out to the size we need.

There is no need to down-res the graphics as they are not "raster based" and as such are perfectly scalable up or down with no loss of quality. In the print world this is done with postscript. In the video world this is vector based graphics.

Does it not work this way on the timelines in Mac-based NLEs? There should be no aliasing whatsoever because the graphics are not a part of the video until render time. Thus I should be able to render a 2K version and a 320x240 without touching the graphics and with no aliasing at all.

Am I just missing something in how the Mac handles this kind of thing?

Brave soul...


Actually, there's lots and lots of things going on. Dynamic rounding, perception of twitter and edge artifacts. I am not going to go there because it's christmas. But suffice to say that the PC world is just as pernickety as this odd little corner of Macdom.

This is not a Mac vs PC or Avid vs FCP issue. I can use FCP to make an SD version of an HD timeline, and with the right boxes checked, it will make a pretty good go at it, but there be dragons! Sharpening, blurring, graphics, twitter of near-horizontal lines, the scaling artifacts of different methods, ooooh, it's a rocky road.

The best thing is that most people don't see all the little give-away tell-tale signs of a less than perfect downconvert. But people who buy for broadcast, or malcontent geeks like us, do and so we hand over to specialist software - which in this case seems to be Compressor.

Trouble is that the PC method of introducing a little bit of gaussian blur to help high contrast edges avoid turning into staircases isn't there in the rubber-glove Compressor interface. So is it there 'anyway'? Compressor reinterlaces movies through a deinterlace command... Does the Detail Levels - used to help edges in UPscales - work in reverse when doing downscales?

Time will tell. I can't - I'm buried until mid February so can't play until then.

So to answer your question, i) Welcome to Macdom, ii) Most of the time, default setting work well but there's better quality to be had for a few mouse clicks, iii) You could do it in FCP (I and a colleague wrote about it in a Ken Stone article), but we think we could do better now. iv) There's always some wretched movie that crops up that simply doesn't look as good as most movies using the standard process, so we have to try anything - I have a couple of patient chickens standing by for these times - to make it work.

Mitchell, I have this entire training video. Larry Jodan's HD->SD workflow is specific to a DVCPRO-HD source!

Definition: DVCPRO HD encodes using 4:2:2 color sampling, compared to 4:2:0 or 4:1:1 for lower-bitrate video formats. DVCPRO HD horizontally compresses recorded images to 960x720 pixels for 720p output, 1280x1080 for 1080/59.94i or 1440x1080 for 1080/50i. This horizontal compression is similar to but more significant than that of other HD formats such as HDCam, HDV, AVCHD and AVCCAM.

In other words, this tutorial is extremely dated and its workflow pre-dates Sony's XDCAM/XDCAM-EX video format. Hence, the FCP worklfow considerations and Compressor dynamics for producing SD from an HD XDCAM-EX source are totally different.

Larry Jordan has since created a more contemporary tutorial entitled "Workflows for FCP Studio 2.0" which includes the HD XDCAM video format delivered by our EX1/EX3 camera's. In this tutorial, you will find a very good "overview" of the considerations needed to achieve a desired output for delivery. However, Larry Jordan is very clear about planning for your final output/delivery. He provides no details on Compressor specifics for "Full HD" to SD conversion. For this Apple's FCP engineers have filled in the details with their Pro Training road trip sessions.

Bottom line: Full HD XDCAM workflows from our EX1/EX3 camera's are at this time, still an emerging/new technologically with very different considerations to make when compared to the traditional "tape based" video formats such as DV/DVCPRO-HD etc. Cheers!

Perrone: You are correct. But....at least in my case I do 90% of my editing in Adobe After Effects. Because I mostly create 30 second television commercials, there are many, many layers, 3D effects, titles, logos, etc....in my projects. My work flow is to render a file (haven't determined the best type yet....ProRes HQ maybe?) out in AfterEffects and then import it into FCP for realtime playback on the timeline. Then I go to tape, dvd, mov, what ever for delivery. So in my case, the graphics will be "flattened" with the video when I import into FCP. I do use a lot of vector graphics (text, logos, etc..) in After Effects, so when I'm in AE, my graphics do scale accordingly. Now on the other hand, when I'm producing a long format project, I use FCP for 80% of the project with AfterEffects graphics sprinkled in the timeline when needed. So even then, many of my graphics, since they came from AE, will be flattened.

Barry: If you re-read my post, you see that the title of his tutorial is "Compressor 3 Essential Training". Compressor 3 came out with Final Cut Studio 2. So I guess my response to your post could be summed up as.......huh? :)

I think you need to watch the tutorial I'm referring to again. He specifically mentions XDCAM numerous times. He even talks about how one format uses square pixels and another doesn't, and what to do in each case. It's definitely current.

Yes we are talking about the same training video, Lynda.com "Compressor 3 Essential Training"

My version is based upon DVCPRO-HD. It is possible that there has been a recent update. In any event, Apple engineers recommend to make the EX-XDCAM HD to SD translation in the FCP timeline and indeed this workflow provides excellent SD DVD results. Cheers!

Use SDHC archive master workflow
 

Maybe someone has tested the following workflow already with good results?

Edit video.

Render final version (HD) to MP4.

Play MP4 out to SDHC Card as archive master via MxR, SDHC, SxS or the likes which, if I remember correctly writes the data onto the card within a complete new BPAV folder combo.

Open that BPAV folder in XDCAM ClipBrowser and save video in DV format.

Sony has a legacy of offering very good downscales.