Has anyone compressed a CF project with CS5 to Standard DVD 720x480. If so what how did it look.
Thanks Bruce

I do that a lot, works fine. I generally prefer to use the scaler in HDLink to make an SD CineForm master, then use CS5 to encode a MPEG2 for DVD.

I do that a lot, works fine. I generally prefer to use the scaler in HDLink to make an SD CineForm master, then use CS5 to encode a MPEG2 for DVD.

I agree.
Particularly if the HD version is interlaced.
When I try to go straight from CFHD 60i to DVD I get more problems with line twitter and other interlacing artifacts. Downscaling to a master 480 60i CF.avi and transcoding from that to DVD usually eliminates all of the problems.

Thanka Dave and Robert,

So do you recomend keeping interlaced all the way? And have you noticed that Encore cs5 does a better job than cs3 converting it to a Mpeg-2

I just got my CS5 system together (16 gigs RAM, 4 disks including 10,000 rpm Velocoraptor for os, 2 TB Raid 0, and another HD) together with the second latest NeoHD and I noticed some differences from my CS3 system for output. With CS3, I would make the sequence for HD Cineform 1920x1080 60i for my AVCHD and then File-Export Movie- and set the outpur format as Cineform and I could then make the output size 720x480 PAR 1.2 and wind up with a widescreen SD file which I would then import into Encore.

In CS5 I would use File - Export - Media and then Set the format to Cineform .avi and set the Video output format to 720x480 and the PAR to Widescreen to create the downscaled .avi for importing into Encore.

1. Is there any advantage to checking the "Use Maximum Render Quality" switch here.

2. Would you recommend exporting the edited HD sequence as 1920x1080 first and then re exporting the "master" you created to 720x480 or just go directly to the 720x480 and not create the master thus saving a step.

3 If you use HDLink, do you use the "convert" panel (above "master" as source file) and then set the prefs to resize to NTSC 16x9

4. Is there any quality or speed difference in scaling down using the Cineform format in Premiere Pro versus using the scaling in HDLink?

I appologize as I haven't had time to work this out myself, but I would appreciate your thoughts since you've already had experience with this.

John Rich

1. If you use color correction within Premeire, then yes. Less so if you use color correction in FirstLight.

2. I prefer to make a 1920x1080 master. If you has spatial filter like blurs, that are correctly computed, but incorrectly so if you do a direct scale output. 1% blur could become a 2.5% blur if you output an HD timeline as SD.

3. Yes.

4. Use HDlink for speed, and a tad more quality (I think -- but subtle.)

Hi Dave what setting do you use in the prefferences when you down size from 19020x1080 to 720x480?
Thanks

Hi Bruce, I wanted to weigh in here, as we have a somewhat different experience.

We have found through many tests with our on-going Tetelestai series, that if we want the very best end quality from our CF'HD (1920x1080p) for a normal 720x480 mpg DVD, we use the following workflow:

We output the full size CF AVI master from an editor (progressive).
We then take the full size master into TMPGEnc (or 2nd choice: Adobe Encoder CS4 -CS5) and convert to standard 720x480 progressive standard MPG, using the highest quality settings in TMPGEnc (VBR, 10bit DC component precision, Highest Motion search precision - with error correction, 8Mbps max bitrate, 2-pass, limited to 8Mbps).

It's very fast (20 fps on older quads, realtime or faster on more powerful) and very nice output. Adobe's encoder seems to work almost as well now, so if you don't have TMPGEnc, it's still pretty much the same. TMPGEnc now handles the CF AVIs very nicely, so no more re-wrapping to MOVs either.

Not all people like progressive to interlaced conversion that some players do for old, interlaced TV's, but we've found it far superior to anything else in almost every circumstance. Most modern players handle the output quite nicely and if it's shown through projectors, or on LCDs or plasmas, it is lovely.

We have found only a few circumstances where our progressive output was a little jerky in a few scenes, but overall the quality is quite outstanding, with great fidelity to the original HD master.

Hi Dave what setting do you use in the prefferences when you down size from 19020x1080 to 720x480?
Thanks

Resize video to NTSC 16x9 and Keep source Aspect ratio checked. Quality set "Film Scan" -- this is overkill but the file is going to be smaller. Always I-frame now.

The reason I prefer the HDLink downscale, it never goes through RGB. It is a 10-bit 4:2:2 YUV pipeline. It handles the 709 to 601 conversion automatically and scales using lanzcos3 YUV 4:2:2 scaler. Having an SD master is nice so you can proof it in progressive 23.976, before the MPEG DVD output which injects pulldown to 60i.

Hi Stephen, I was almost at the same settings as you were using. What target bitrate do you use?
Dave I am going to try that 2.
1question don't either adobe or tmpegnc compress more or should I say screw things up converting to mpeg-2 or does it just convert it.

Resize video to NTSC 16x9 and Keep source Aspect ratio checked. Quality set "Film Scan" -- this is overkill but the file is going to be smaller. Always I-frame now.

The reason I prefer the HDLink downscale, it never goes through RGB. It is a 10-bit 4:2:2 YUV pipeline. It handles the 709 to 601 conversion automatically and scales using lanzcos3 YUV 4:2:2 scaler. Having an SD master is nice so you can proof it in progressive 23.976, before the MPEG DVD output which injects pulldown to 60i.

Dave do you mean check the iframes only box

Hi Stephen, I was almost at the same settings as you were using. What target bitrate do you use?
Dave I am going to try that 2.
1question don't either adobe or tmpegnc compress more or should I say screw things up converting to mpeg-2 or does it just convert it.

There's not much gain (and can be problems) trying to push the standard DVD spec beyond it's max, so we just use the standard 8 Mbps max for video and uncompressed 1536 kb/s PCM for audio, so it adds up to the 9707kb/s max. Seems to be less problems and better that way.

This is the highest quality setting we've ever found for standard DVD's. Compression engines for both prog's seem to be pretty close now, though I still find TMPGEnc a tiny bit better overall. Probably not enough diff to warrant buying TMPGEnc, but worth using if you have it.

If quality is an issue and space is still at a premium, you can go Dolby, but otherwise it's the best overall for the old DVD standard.

I have also done 35Mbps BRMV output from the masters, but have not really done extensive testing. It looks good compared to the CF masters, with just a bit of artifacting at times.

Dave do you mean check the iframes only box

Yes, but I'm doing a lot of 3D stuff lately, the primary reason to use I-frames. Most use cases the non-I frame sequence is perfect (saving you 15-25%)

David,

Here we are in 2011 just about. Has this workflow that you recommend changed any in the last several months since you posted this? My system is CS5, 64-bit with Cineform Neo3D. I use Encore for the DVD rip and my clips are all CFHD 1920x1080p with a lot of Cineform colour correction in the Premiere CS5 timeline. I'm downscaling to DVD widescreen.

At the moment, I've tried importing the Premiere timeline directly into Encore. It's a 2 hour show and the transcoding has been going 12 hours already and still isn't half way (8GB RAM, quad core 64-bit system).

To clarify for best PAL results, is the workflow:

1. Export the full size (1920x1080p) CFHD master from Premiere CS5 (.mov or .avi?)
2. Create an SD master using HDLink to downscale the resulting .AVI or .MOV to 720x576 (for PAL) - Don't know if there's a selection for widescreen here, or HDLink does that automatically.
3. Open encore and import the 720x576 SD master for output to DVD.

I don't know what I'm doing wrong, but whenever I try to export a 2 hour 1920x1080 movie that is CFHD with cineform colour correction, premiere gives me an estimate of 11 hours to export this. Is there any way to make this faster? I only have 2 video tracks and 2 audio tracks.

Many thanks for any help,

Kris,

FirstLight color corrections are faster than the Premiere filters. The biggest issues is the export doesn't thread very well. However your exports still seem very slow. Please contract support after the holidays.

The biggest issues is the export doesn't thread very well.
Will this biggest issue ever change? :)

There has been many CSNext's/CFNext's now without this issue being resolved, hence the question. I see very close to the same same export times on a HP Z800 with 2 x six core X5670's that i had on a HP xw8600 with 2 x four core E5450's. So more cores does almost nothing today, with CF exports that is. Fast RAID or no RAID makes a very minor difference.

It seems to be an issue with decoding CF as well since exports *from* CF takes approx three times longer than the so-hard-to-decode-AVCHD-codec takes. One hour native AVCHD exports to MPEG2-DVD @ 20 minutes with the Z800 while the same CFHD footage exports to MPEG2-DVD @ one hour. Thats why i use native AVCHD in all my edits today, because it saves tons of time. (Yes, i am a nerd that test this...)

On Topic:
I use the CF codec to archive footage and render out one HD master and one SD master from the same Timeline and dont use HDLink to downscale the HD master because HDLink uses the wrong PAR and thus stretches the SD output. I have compared the quality between HDLink downscaling and think the CF Exporter in Premiere Pro does a better job when downscaling the footage when it comes to quality, PAR issues exluded.

Sorry for the semi off topic post.

Hoping for faster exports from Pr and the ability to choose between current PAR's and correct PAR's in HDLink in the future! :)

Speed is Premiere API issue, we have the fight with Adobe every so often, as we don't have it anywhere else in any other tool. We have temporarily disabled some optimizations, as we found 5.0.3 less stable forcing us to back off on more threading. We are working with Adobe, so I hope to one day work out the magic combination. This has been very frustrating seen the mess of CS4, forced us to find more business on other NLE platforms.

I think you might be wrong on the PAR issue, otherwise we can fix a bug. We don't put the PAR in the headers AVI headers using HDLink, but the media is scaled correctly, and can be used as-is with interpret footage. Where have you found this not for be true? Why not put PAR In the header? No idea, it seemed many AVI many tools don't use it as AVIs are so old they pre-date the concept of PAR. Fortunately HD is mostly square pixel, and more often progressive these days.

Speed is Premiere API issue, we have the fight with Adobe every so often, as we don't have it anywhere else in any other tool. We have temporarily disabled some optimizations, as we found 5.0.3 less stable forcing us to back off on more threading. We are working with Adobe, so I hope to one day work out the magic combination. This has been very frustrating seen the mess of CS4, forced us to find more business on other NLE platforms.
O.k, i hope this will change soon. :)

Where have you found this not for be true?
I was wrong about the PAR in the file header, but i was talking about that Adobe changed their PAR's in CS4 so HD downscales correctly to SD, IE that the overall geometry of the image is not distorted. Images downscaled with HDLink downscales the same way pre CS4 did, ie it stretches the SD image slightly.

To see it:
1. Create a HD Timeline in CS5 and export it to SD.
2. Export the HD Timeline to HD and downscale it in HDLink.
3. Import both SD files and place them abowe each other on a SD Wide Timeline
4. Toggle Video 2 off and on and compare the overall geometry between the video

Result:
The video downconverted with HDLink is stretched horizontaly while the overall geometry is untouched in the footage CS5 exported.

Additional info:
Yes, there are black bars in the SD footage CS5 exported, but it can be cropped away prior to export so the video indeed fills the whole frame.

If i had that option to downscale correctly in HDLink i would use it everytime! :)

I've just done your test. This is regarding whether the 16x9 display display is represented by 702x576 or 720x576. While seems the recent consensus has moved to 702 (Adobe changed at CS4), yet seems tools still do there own thing.

1920x1080 is exactly 16x9 square pixel.
You want to map all these pixel to SD at 720x576. CS5 only fills to 702, Vegas fill to 720 but used the "new" pixel aspect, HDLink fills to 720 but uses the old pixel aspect. The Vegas approach loses data on the top and bottom, the Adobe approach doesn't fill 9 pixels on the left and right edges, the HDlink approach doesn't lose any pixels anyway, but aspect could be different by 2%.

We can easily change, but to which method? I think I prefer the Vegas approach.

We can easily change, but to which method? I think I prefer the Vegas approach.
I think i prefer the Vegas method as well, but would love to have the ability to choose between "HDlink approach" or "Adobe approach" or "Vegas approach". :)

HDlink approach = The way HDLink does HD > SD today.
Adobe approach = The way Adobe does HD > SD today.
Vegas approach = The way Vegas does HD > SD today.

However, if there can be "only" one choice i prefer the Vegas approach, IE the same result one get if one Crop the output, top and bottom, in the Export Settings dialog in Pr CS4/CS5.

The Vegas approach is easiest.

http://cineform.com/downloads/CFVideoChangePARFIX.zip

Unzip and replace in C:\Program Files (x86)\Common Files\CineForm

Now HDLink HD 16x9 to SD 16x9 (PAL or NTSC) will use the newer pixel aspect ratio.


Hint: if you want to make SD DVD fast from a HD CineForm AVI or MOV master. In HDLink select Fast Scale, for NTSC which is less than half HALF resolution is either direction, the fast mode tells the CineForm to decoder only output 960x540 and scale from there. Quality is still excellent. I converted a 90 minute HD project to SD in under 30 minutes -- the SD AVI ro MPEG2 DVD is super fast now.

The Vegas approach is easiest.

http://cineform.com/downloads/CFVideoChangePARFIX.zip

Unzip and replace in C:\Program Files (x86)\Common Files\CineForm

Now HDLink HD 16x9 to SD 16x9 (PAL or NTSC) will use the newer pixel aspect ratio.
Hurray! :)

Thank you very much, Sir!

I just tested it with a long clip and i love the workflow with render a HD master from Pr and use that HD master and let HDLink downconvert it. If the SD version looks good, the HD master looks good as well. This will save me tons of time.

And, if i import the downconverted file into Pr the Generating Peak file process is *lightning* fast compared to the files exported via the CF AVI Exporter from Pr. That fix is appreciated as well and i hope it will find its way into the Pr CF AVI Exporter in the future! :)


Hint: if you want to make SD DVD fast from a HD CineForm AVI or MOV master. In HDLink select Fast Scale, for NTSC which is less than half HALF resolution is either direction, the fast mode tells the CineForm to decoder only output 960x540 and scale from there. Quality is still excellent. I converted a 90 minute HD project to SD in under 30 minutes -- the SD AVI ro MPEG2 DVD is super fast now.
Great, i will try that! Thats great to know when one just want to make a DVD for rewiev. :)

Here is the fix for the AVI export having fast to import audio.

http://cineform.com/downloads/CFRenderProc.zip

Unzip and replace in
C:\Program Files\Adobe\Common\Plug-ins\CS5\MediaCore\CineForm
and
C:\Program Files\Adobe\Adobe Premiere Pro CS5\Plug-ins\Common\CineForm

Let me know whether that is better for you.

Let me know whether that is better for you.
Wow, thats fantastic! :)

I rendered out one clip with the old CFRenderProc.exe and one clip with the new CFRenderProc.exe and imported each clip into Pr CS5 and the difference is huge:
- Old CFRenderProc.exe = Generating Peak File took 91 seconds
- New CFRenderProc.exe = Generating Peak File took 8 seconds (Yes, eight...)

Each clip was ten minutes long. This fix makes the rendered clips file size a tad smaller than before and though i did not time the render times i got the feeling that it rendered out from Pr faster than before. I just changed .exe to .OLD, so i will switch back and test later on a longer Timeline.

Great job CineForm! It is indeed appreciated! :)

Edit:
Yes, the rendering times are indeed improved with this fix. I renderd out the ten minute Timeline to test, and with the old CFRenderProc.exe it takes 8 minutes and 12 seconds and with the new CFRenderProc.exe the same Timeline takes 6 minutes and 41 seconds. The new CFRenderProc.exe renders out my Pr CS5 Timeline approx 18,5% faster than before. :)

I rendered out native AVCHD 1080i clips from a AVCHD 1080i Timeline > CFHD 1080i.