Ok thanks for that. I will try it later today.

OK it installed perfectly. Stuck a 720P50 clip on the FCP timeline, went export, Quicktime Conversion, chose DNxHD codec, and it made the same sized file. A 20 second clip came out at 283Mb in size.


But. The clip looks washed out. The contrast is shot. However I have chosen one of the other Avid codecs and got stunning results downconverted to Pal DV. I am now importing that into compressor and out with Cinema craft Encoder MP.

I've been using PhotoJPEG at 80-100% (I believe 100% is 4:4:4 RGB and virtually lossless) for over a decade, mostly for archiving motion graphics final renders, though I also used it for downsampled HDV and DV50 work. With more compression (75% to 85%), it makes for a great real-time playback format in Apple's Keynote - especially 1280x720.

Great legacy format.

I realise it, Dominik.
You proved your method to me and you are obviously an expert in this field but sadly, despite your best efforts, some others don't seem to want to learn from you.

I hope you won't give up.
People like you are just what these forums need.

Can't help, but you sound a bit ironic to me. ;)
This one was posted by Perrone done with lanczos in virtualdub:
http://www.dvinfo.net/conf/attachmen...test_small.png
This one was posted by me done with Compressor:
http://www.dvinfo.net/conf/attachmen...compressor.png
They are almost 100% identical. Hence Compressor is doing something equal to lanczos.
HuffYUV can do RGB, but it is a lossless codec, so it produces very large files. About 50-60% file-size of uncompressed, similar to rar-compression.
That sounds very interesting, especially for pre-rendering of green-screen work because of the alpha-channel-capability. Earlier I used HuffYUV for that, now I've got a faster computer. ;)

Yes, there are some other great intermediate codecs, which are capable of 4:4:4YUV/RGB. They're especially useful for special-effects-work. But I think that with 4:4:4 you don't gain any advantage for pure editing/grading, because the end-consumer/release-formats are at most 4:2:2, so using 4:4:4 is just a waste of harddisk-space.

The codec has a number of options. One preset (DNxHD 36) is specifically for intermmediate work. I am about to try it now.

More info here:

Avid DNxHD Codec

Cool article about Iron Man being cut in this HD proxy Format:

DNxHD Marvel | Behind the DNxHD 36 workflow for Iron Man

Ok I have had a few plays with it now and it does give off a very good video file which I boought into Compressor and then out again via Cinema craft Encoder MP. I did a 10 pass VBR with a max of 9000, min of 2000 and average of 6000.

Quite a good clip to use as it has plenty going on with all that confetti. Plus the bride ain't bad looking either!

Original untouched file is here. Shot with EX1.
Right mouse click and "save target as"
http://www.steveshovlar.com/cinemacr...44_1441_01.mov

The finished file is here.
Right mouse click and "save target as"
http://www.steveshovlar.com/cinemacr...tEncoderMP.m2v

Can anyone improve on that finished file?

I tried DNxHD and my impressions are:
-it is 4:2:2, not 4:4:4
-the 36mbit/s-variant has much more artifacts than 35mbit/s-XDCAM-EX
-at all bitrates (and without alpha-channel) it needs for playback about 2 times as much CPU than ProResHQ and about 3 times as much CPU than XDCAM-EX
-it doesn't support 1080p/29.97

So I can't see any advantages using it except the alpha-channel-capability.

Maybe something got lost in translation.
(I was saying you do a good job - I meant it).

Sorry Dominik, I am listening, but I'm not understanding. You (and others in this thread) talk over my head quite a bit.

So what your saying is that because I compressed to standard DV codec (YUV), it's going to look crappy. Sorry but this doesn't make any sense to me, because the commercial was originally produced and broadcast in DV codec. I rendered the project in AfterEffects directly to a DV codec file (Quicktime mov). I've seen it on the air and it looks fine. By my logic this tells me that it should be possible to convert HD to SD - DV codec, but that the tools I'm using (Compressor) arn't working very well.

Compressor does offer conversion to DVCPRO50. Isn't that a 4:2:2 codec? I can't remember. I think it also will convert to the Uncompressed codec.

My problem is all the local broadcasters in my market broadcast their commercials using the DV codec. When I send them copies of my work (dubs) I have to send them as Standard Def DV codec mov files. Another option is to send them as a DV Stream, but that's basically the same thing as a DV codec mov.

On January 5th (when all our new equipment arrives) I'm especially looking forward to installing our new AJA Io HD device. Supposedly I does a great job converting HD to SD using hardware based conversion. Once it's all hooked up, I will have an SD monitor connected alongside our HD monitor so I will be able to instantly see what an AJA IoHD SD conversion looks like. But I agree with you, I not very optimistic that it will look better than the best software based conversion.

Still looking for a simple HD to SD (DV codec) solution on a Mac..... (hopefully you now understand why I keep wanting to transcode to DV)

Neither is ProRes, DVCProHD, HDCam, Etc. Other than an image format like PNG, JPEG2000, or something similar, you are not going to get 4:4:4. Most people can't aquire in 4:4:4, but it does have some uses for VFX and such. If that's your bag, then this isn't for you.

Not in my testing. What was your source? I tested mine against some 2k RED footage and it was very, very nice.

I can't verify this because I am not on a Mac. So I'll take your word for it. The closest I can come to ProResHQ on the PC is Cineform, and I can't use it in my new NLE, which is why I went on a hunt for a new codec. I am trying to be respectful to the fact that this is a Mac thread and not a PC one though.


This surprised me until I realized what the codec was for. Film. This is not a broadcast codec. It is there to support 24p, 25p, and IVTC film rates. Honestly, if you're going to NTSC broadcast, you'd want to follow standard, and I don't remember 1080p being in the rec.709 standard. Though maybe it is now.

Advantages over what though? The advantages of working in uncompressed are clear. The Advantages over working in ProRes, I can't speak to. How do the file sizes compare? For us poor PC folk, it has an advantage over nearly any other choice we have except maybe Cineform, but this is free.

It was not my intent to come to a Mac thread and convince everyone to use an Avid codec! :) It works well for my purposes, and I think others may find use for it. One of the PRIMARY advantages to me, is the ability to move high quality proxy and master files from Mac <-> PC without gamma shift or other issues. ProRes falls down badly there, and the 4:4:4 codecs are HUGE in comparison.

For me, and as noted in the linked articles, the ability to cut HD sized proxies gives great advantage. I can preview for clients in HD and view my cuts and grades on my laptop. This was not something I could easily do before. I recut my latest project masters over night. Being able to open my masters and play them in near realtime on this underpowered laptop is amazing. I wish AVCHD was as easy!

Sounds like a gamma conversion issue. This is one thing I wish the Mac and the PC could sort out. Not sure why you saw it in one codec and not the other.

Try this - I downconverted with Compressor
http://www.workingsheepdog.co.uk/vid...or-MPEG-2.tiff

And here's a still from your CCE file:
http://www.workingsheepdog.co.uk/vid...EncoderMP.tiff


Look closely at the bride's necklace.

I'm sorry I did a mistake. What I wrote is true for 4:2:0 but not for 4:1:1, because 4:1:1 has quartered horizontal chroma-resolution, but FULL vertical chroma-resolution, so NTSC-DVs chroma-subsampling doesn't produce an interlaced look on strong colored edges.
But there are two issues with QuickTime-Player:
-DV-Movies will initially be displayed with very bad quality. You have to check the "high-quality"-checkbox in the movie-settings to get better quality.
-then the quicktime-player shows indeed a full-resolution luma-channel, but always deinterlaced chroma-channels. I don't know how to turn that chroma-deinterlacing off.

I only can access the full-resolution DV-file, if I load it into AfterEffects. Then I see that your downscaling hasn't worked well, because there are jaggy edges.
I don't know what went wrong, but if I try it myself to do that HD->DV-conversion in Compressor, then everything looks fine. I attached examples to illustrate:
The first one is a still of your dv-file. The second is a still of my dv-result.

Mitchell, this article may help. It's very clear, speaks on the variety of codecs out there, and what they do (in non-technical terms):

http://www.avid.com/resources/articl...out_Codecs.pdf

I think Dominik, and myself to a degree, get in these semantic arguments for no purpose. The primary purpose of a codec for editing, is to allow you to do your work while losing as little information as possible. In the camera, what we WANT is high resolution (4k is great, 8k is better) and no color subsampling, so 4:4:4, and we want as much fine granularity as possible in that color, so 12 or 14 bits please! Now I shouldn't have to tell you that saving 4k, 4:4:4, 12 bit data is going to make data files that are astronomical. So choices need to be made on what to throw away. In the camera, we generally see resolution subsampling so that lovely image is now 1920x1080. And we lose some color usually, so that goes down to 4:2:2 or even 4:2:0 in the case of the Sony XDCamEX. And we record 8 or 10 bits instead of 10 or 12. More expensive cameras throw away less data.

When we get back to the NLE, our camera has tossed away a ton of stuff already, so we'd like to preserve whatever is left if possible! In film, the norm has been to use uncompressed data which preserves all we can. Problem is, it makes very large files which require big and fast disks to handle properly. So we have to make more compromises. But it's here that we have some choices. If data is going to the web, do we really NEED massive files when we're going to lose 80% going to the web? If we are going to broadcast and we know that 50% of it will be tossed, do we really need the pain of 20 hour renders?

So you can chose two things. Your editing codec, and your mastering codec. I typically chose to master to uncompressed because I could always go back later and re-render to anything lower. And for editing, I chose to use uncompressed SD, which was still big files, but not as big as HD. You can use ProRes for both mastering and editing. I changed to Cineform to do the same. The Avid codecs don't fit here. They fit in the idea of proxy and seperate master. Not a one size fits all. It's simply a choice. However, of all the codecs out there, the DV codec typically is the most lossy of all the common ones. Coming from uncompressed 1080p, it probably loses 80% of the information, or thereabouts. It is absolutely the worst thing to do to a file. So do NOT go there until the very end of the workflow when you are producing a deliverable.

DVCPro50 is nice, but not full raster. So you'll take your 1920x1080 and shring it to 1280x1080. better than DV, but still a terrible thing to do if you don't have to.

If this is what you have to do, then it's what you have to do. BUT, don't do this until the VERY END of the workflow. Render to DV, and master to tape or whatever. Save it for the very end.

The very same Aja codec is available as a free software codec. I've used it. Honestly, I like the Avid one better.

I think we've inched closer though. Especially given Dominic's success with his Compressor results, and my posted resuts with Virtualdub that are available to Mac users running bootcamp.

I know, but I guessed that you implied that DNxHD can do 4:4:4 by saying that it can do RGB:
I used XDCAM-EX-source for testing. I attached an 100%-crop-example. The left side is avid with 36mbit/s, the right unaltered XDCAM. There are clearly more artifacts on the left side.
advantages over using ProRes.
very similar.
Btw, watch the graph at the top on page 10 in this document. ;)

Andy, remember although the CCE file is out of compressor, it was also taken out of FCP using quicktime conversion into the Avid codec.

I'll run two tests. One from FCP to compressor using CCE, and the other from FCP via Prores422, then onto a SD FCP time line and out via compressor with CCE.

I'll post screengrabs when done.

I don't have much time to talk right now (I will later tonight).

But I'm amazed at how Dominik has made my DV footage look so much better. I'm wondering if it's how you guys are taking "screen shots". Are you literally taking a screen grab (on a Mac = shift, command, 4) or are you exporting a frame from Quicktime?

Dominik says that he could only download my crappy DV conversion. But he has taken crap and made it look great! This makes no sense to me.

Sorry guys if I'm slow on the uptake. But when the light bulb finally comes on it will be much appreciated.

BTW, Dominik I knew about the high-quality thing in Quicktime. I'm guessing that trips a lot of people up though. When do you think Apple is going to remove that check box.....it's about time. :)

No, I downloaded your dv-version and your png-compressed HD-version. The left attachment was a grab of your dv-version, the right was a grab of my dv-version made out of your hd-version with compressor.

I used AfterEffects for the grabs.
It would suffice if it would be turned on by default. ;)

No, only implying that the codec has an RGB option built into the choices. My guess is that it's there for Alpha Channel support.

That's weird. I am going to compare some 1080p high motion stuff I shot 2 weeks ago (flowing water) and see what happens. I'll post a difference still as well. Source to uncompressed, source to DNxHD, and Uncompressed to DNxHD.

Prores is pretty darn good. But I can't use it, and neither can any PC person. So for collaboration, it stinks. It also has no alpha channel.

LOL! Check out the paragraph BEFORE the chart: "The chart below plots the PSNR value for each image frame in the Digital Cinema
Initiatives StEM (Standard Evaluation Material) sequence. The HD version of this
sequence was converted to a 10-bit, Rec. 709–compliant sequence."

i.e. ProRes couldn't handle the RGB values from the original file so we truncated it to something we could handle, and then compared ourselves to our competitors! And even WITH that the gap from DNxHD to ProRes was 2db. Visually indistinguishable. Mind you, this is ProResHQ which most people don't use but probably should for this work.\

Gotta love Applespeak!

If the codec would support RGB, then it would support 4:4:4, because RGB and YUV4:4:4 are effectively the same. But I can't find a RGB-Option in the dialog:

Really?!! That's great news! When you have the time, can you pretty please tell me exactly what setting you used in Compressor. (remember to dumb it down for us non-technical guys, hehehehe)

Thank you very much Domink.

Perrone, I think your solution is great too, it's just that Dominik has seemed to have solved it using the tools I already have. :)

Completely understand! And that's why we are here. To find Mac solutions for Mac users. My solution is meant as an alternative at best, either for those without compressor, or those who, for whatever reason, can't get the results needed from their Mac based solution.

Mine does have the benefit of being free, so for those who have bootcamp but don't have Compressor, there is a way.

I hope Dominik gets you a screenshot, because this should solve your problem. And honestly, should be a sticky in a new, clean thread. That is, if we've actually solved the HD > SD conversion for Mac folks. And Dominik's posts and screenshots lead me to believe that he has.

a sticky please, Chris
 

Gentlemen, I think this thread shouldbe turned into a sticky.
What a wealth of information :)

OK gormless question. How do you get a tif from an M2V file? I opened in Quicktime, go export, buut there is no export frame. I tried export image sequesnce and have spent 20 minutes clearing 1000 tifs from my desktop. Changed the amoount per second to 0.5 to get a few, but the quality is not as good as the .mov.

All I need is to find the frame I like, and then export that frame as a tif.

Ok,

So here is a screen shot of my editing space. Essentially, I took 5 seconds of a closeup of water on a fountain. So organic movement, and unpredictable by the codec for the most part. I stacked the original, and uncompressed, a jpeg2000, and the DNxHD175 on the timeline.

I then did a difference composite for each one. I have highlighted the Avid against the original source. I have histograms (Luma, R,G,B), Waveform, Vectorscope, and RGB Parade open as well as the preview window.

You can see from this that there is nearly NOTHING. For all intents and purposes, DNxHD at this bit rate is lossless when presented with an XDCamEX image.

[Edit]
Oh, file sizes for 5 seconds of 1080/24p:

Uncompressed: 972,008
Jpeg2000: 114,882
DnxHD 175: 107,522

Interesting results, Perrone, thanks for posting.

OT: I notice you're from Tallahassee. Have you done any shooting down at Wakulla Springs? I lived there as a kid.

Ok here's the two TIFs I made.

First one is using Cinema Craft Encoder MP, 15 pass VBR.

http://www.steveshovlar.com/cinemacraft_encode/CCE_.tif

Second one is using prores422 into Compressor then out using Cinema Craft Encoder MP, 15 pass VBR.

http://www.steveshovlar.com/cinemacr...tprores422.tif

Pretty sure I have captured the wrong way. The tifs don't looks as smooth as the video.

I didn't do anything more than I showed in this video except for selecting QuickTime/DV as output-format:
YouTube - high-quality HD to SD-DVD conversion with FCP
You can get PNGs with MPEG Streamclip. PNG is better than TIFF for web-use, because it's lossless compressed (vs. uncompressed TIFF) and well supported by web-browsers.
I was talking about the 36mbit/s-variant not 175mbit/s. The 175mbit/s-variant will of course look very good.
I attached a screen-shot to show you what I mean.

Steve you've done something very wrong there my friend!

I completely misundderstood you! So for grins, I'll try a 36mbps render now.

New difference attached...

File size: 22,082

Sorry to be an absolute ANORAK over this but...
 

If I am not much mistaken, DVCPro50 is Panasonic's souped up version of DVCPro (Panny's 'DVCAM'), which uses 2.5:1 compression rather than 5:1, therefore getting pret-ty close to DigiBeta. I used it a lot for sucking in and laying off to DigiBeta/BetaSP with my Aja IO. Useful format for stuff with lots of motion graphics 'pre-ProRes'.

I believe you were referring to DVCPro-HD? Or more specifically the 720p variant at 50 mbps? If so, would that be 960x720? ... Does anyone care?

With that and a penchant for PhotoJPEG, I feel like a 'Dead Codec Preservation Society' - "Anoraks Up!"

Yep. Always forget the SD variant because I never used it. Went from VHS to SVHS to DV. I trade DVCam with the local PBS folks and that's about the extent of it. No Digibeta for me! Couldn't afford the VTRs!

O Captain my Captain! LOL!

OK last night I completely messed up the images of the clips I was trying to show. Downloaded and installed Mpegstreamclip and it was very easy to grab a frame.

Here are the results of using Compressor with a two pass VBR, and using Compressor plugin Cinema Craft Encoder MP.

Now this is how I got the footage to Compressor.

I dropped the 720P50 footage on its own timeline, then exported it to prores422 720P50, self contained movie.

I then made a new sd timeline , PAL and dropped the prores422 file onto it. FCP asked if I should change the timeline. No is the answer. Then I rendered the timeline. I now have a nice SD timeline with the 720P50 footage converted down to SD PAL. It looks lovely and clean, no aliasing.

I then export this timeline to Compressor, and did a two pass VBR on the 90 minute DVD setting, and then did it again using Cinema Craft Encoder MP, do a 10 pass VBR.

First thumb is the original footage, second is compressor, and third is Cinema Craft Encoder MP.

http://www.steveshovlar.com/cinemacr...e/original.png
http://www.steveshovlar.com/cinemacr...compressor.png
http://www.steveshovlar.com/cinemacraft_encode/CCE1.png

Thoughts?

I tested this method and I figured out that this method gives the same result as using Compressors "better" rescaling method, but much faster than doing it with Compressor. The quality is great, but if you want the absolute highest quality and you have enough time then use Compressor with its "best" rescaling.
You should really check the detail-settings in your ex1. The sharpening looks really ugly. I always recommend to turn the detail-function entirely off.

Btw, you're not converting to interlaced SD, so you're effectively converting 50fps to 25fps. Do you intend to do that? If yes, then I would recommend not to shoot with 720p/50 but with 1080p/25.

It is not necessary to to export to anything. Just drop the HD sequence on the DV timeline. Did you test that workflow?

The CCE1 comes out a little better than compressor.

I agree. Cinema Craft Encoder MP is better than Compressor. But then it should be for 800 bucks.

I haven't but I will certainly try this out.
Thanks
Steve

Good point on the detail. As I shoot 95% weddings the client mostly wants the video look, so that's what they get. I wil shoot some footage tomorrow morning and compare detail on and off.

I shoot 720P50 because its much better for slo mo in the snogging musical interlude. 50% of 50 frames = 25P.