DVD Rendering From HD Still a Crap Shoot?

[Harry Simpson]

Two years ago if I captured 1080p with the 5D mk2 it was a nightmare to create a SD DVD disk - first it looked terrible unless you followed a cookbook of steps which involved using VirtualDub and re rendering afterward etc.
With Sony Pro 10 or Studio 10 Platinum is it any easier? Or is it still a task to get a great looking DVD from HD capture?

[Leslie Wand]

it's still a pita - and probably will be evermore since you're trying to get a pint in a half pint pot....

[Harry Simpson]

awww shucks - was hoping technology had solved this in the last couple of years. I've got my recipe! ;-)

[Roy Alexander]

I'm not an expert but I do know about sharp looking DVD's. I can honestly say that up to now, using Sony VMS platinum HD 10. I have edited HDV videos and using DVD Artitecht have burnt DVD's from them and when playing the DVD's back using an upscaling player and a HD TV have been very suprised how good the image was. In fact viewers have remarked that there was very little difference in the quality between the SD DVD's and the blu-ray discs. Best part it only took about 30 mins to burn a 10 min. video.

[Randall Leong]

Roy,

That's only because you are shooting 1440x1080 HDV which has a Pixel Aspect Ratio (PAR) of 1.3333:1 - close to that of 16:9 widescreen 576i PAL SD. The 5D Mk2, on the other hand, shoots in 1920x1080 with a 1:1 (square pixel) PAR. That PAR is too far apart from the 1.4545:1 ratio used in 576i PAL widescreen or even the 1.2121:1 PAR of 480i NTSC widescreen. Vegas' own encoder and those of most other NLEs by themselves do a poor job of resizing square-pixel material to a format that uses non-square pixels. Severe artifacts will occur in this type of conversion simply because the NLEs' own encoders do not resize the pixel shape correctly.

[Roy Alexander]

Randall. I take on board what you say, and acknowledge your expertise. However, being in the movie business for 60 odd years, I believe, that although quality of the image is important, the actual content of the film or video is more important.

[Chris Harding]

Hi Randall

Actually my older HMC70 cameras used to produce an AVCHD file of 1440x1080 with the usual 1.333 PAR... the newer cameras now shoot 1920x1080 with square pixels so the end result despite the "bigger" picture is a bit poorer!!!

Pity there wasn't a solution that was simple!!! Like Roy I like to rely on content as there is no real "one NLE" solution

Chris

[Chris Harding]

Hi Guys

I actually decided to do a short test with 1920x1080i video using different conversion formats and then checked the results on a DVD PAL Widescreen

All clips were rendered to Main Concept MPEG2 using the DVD PAL Template (16:9, 25fps, lower field first and a variable bitrate...default settings)

(1) Clip 1 - I rendered the Panasonic MTS file direct from the camera ..chose the correct properties in Vegas 9 and then also checked the de-interlace method as interpolate.

(2) Clip 2 - I transcoded the MTS file from the camera with MainConcept's Transcoder from Panasonic directly down to Widescreen AVI 720x576 and chose the correct properties and rendered the same as the above.

(3) Clip 3 - I converted the MTS file in NewBlue Upshift to HDV but still 1920x1080 at 50mbps but also told the program to make the file progressive so I could leave the de-interlace setting in Vegas to none....chose the properties Vegas wanted and again rendered the same as above.

(4) Clip 4 - I converted the MTS file in Cineform to their huge AVI file ...in Vegas again I let Vegas decide on the file properties but set the de-interlacing again and rendered as before.

OK all were written to a DVD and played on my 32" HD TV and watched several times....First of all my wife could not tell the difference between them but if I looked pretty hard I could see a tiny loss in definition on the clip that was converted to SD before rendering....it was very hard to spot and unless someone was looking for it they wouldn't have even seen it.

Just for interest I use Upshift purely because my little Duocore cannot handle big AVCHD files so they are converted to HDV but made progressive and also kept at 1920x1080.

Has anyone else done a test like this???? Apart from filming a perfectly lit resolution chart do you really think that the average person can actually tell any difference in something like a wedding where the content is being watched rather than technical issues ??????

I would gladly welcome any suggestions that have a VISIBLE difference on a DVD !!! Sure the image is absolutely stunning during preview as you are looking at 1920x1080 images not 1024x 576 ones ....however I have yet to see any "method" of HD to SD that sweeps away the rest....

Please tell me if you have anything that does make SD look like HD and is on a normal DVD!!!

Chris

[Christopher Young]

Have been battling these issues for years now and finally have found something that works pretty well with 1920x1080i footage from XDCam HD, EX1's etc. The SD widescreen DVD's I am now making are by far and away the best I have made to date from HD 50-mbit 422 MXF files. 35-mbit 420 files also work well.

The software is Mainconcept's 'Reference', not cheap, but even picky clients have been happy with the results. Try the demo download. It's watermarked but well worth a try. I'm PAL based so use the MPEG-2 DVD preset with these alterations. Check the '2 Pass' box and under PQ select '31 Slowest/Best' [default is 16 Balanced]. Under Advanced Settings set Target Bitrate to 8000 [default is 6000]. Leave Maximum bitrate at 9500. Under Auto GOP Placement select 'SCD Fast' [default is SCD Refined]. This forces a new GOP strat on scene changes. Lastly set Noise Sensitivity to '1'. If you have good clean footage this will give you the cleanest encode. Try these as a starting point and see how you go;

[Mike Kujbida]

Originally Posted by Christopher Young

The software is Mainconcept's 'Reference', not cheap...

Chris, if you mean Reference Engine: MainConcept, you weren't kidding when you said "not cheap".

[Ron Evans]

I find TMPGenc T4 or T5 gives the best downconvert and encode for DVD. I mainly use Edius to edit so then would export a HQ HD file for TMPGenc to use but in Vegas a lossless HD export also works in the same way for the1920x1080 AVCHD files I have now. I also use 2 pass VBR with 10 bit precision and highest motion detection. With playback on upscaling DVD player the picture is almost as good as the Bluray. You can see the difference in fine detail in large sections of colour but the average viewer would not see much difference. This approach is a lot better than Vegas or Edius encode from their timelines.

Ron Evans

[Chris Harding]

Thanks Chris

I guess the trial version is worth a go if it's that good!!!! Wow!! I see what you mean Mike...get a fairly good camera for that price too!!!

Excuse the ignorance but would one export the timeline from Vegas as an MXF file??? and then encode to MPEG2 in reference.

It is a huge pity to edit in HD with pristine footage only to see the end result on DVD not the same quality!

It would be so easy to just burn a BD for my brides but players in the West seem to be scarce in homes!!

Chris

[Phil Lee]

Hi

You can achieve something similar to MainConcept using Sony Vegas. MainConcept reference design has a de-twitter filter and possibly a blur for interlaced sources that gets applied during the resizing, the actual MPEG encoding engine in the Reference design is exactly the same as the MainConcept encoder in Vegas, don't waste your cash on it. I find the MainConcept reference rather lacking in detail with a fuzzyness about it.

Try this, from the Sony timeline, export your footage as Sony YUV using the PAL DVD Widescreen pixel ratio settings, make sure under project settings you have selected "Best". Now bring this back to a new project, right click the clip and tick "Reduce Interlace Flicker", then export it using the DVD encoder, up the average bit rate to say 9000, all quality settings to high, and make sure the project settings are set to Best. See how that looks, this will give you an output very similar to the MainConcept one.

After lots of messing around I find there are 2 extremes of HD to SD. One is you have sharp footage but this will show interlacing artefacts/stepping/combing over fine detail, plus some line twitter over horizontals. Or you can have footage without artefacts/stepping/combing with next to no line twitter, but the footage will have much less resolution. You can't have both. Everyone's different methods simply move the slider in between these two extremes to suit their own tastes and types of footage.

For viewing on a computer, the sharper with artefacts end of the scale looks worse as the artefacts show up very clearly, but on a TV, when sat back watching, those artefacts are better handled by most TV de-interlacers plus being sat back you hardly notice them, but you can see the extra resolution obtained by that method. The Mainconcept type approach produces fuzzy but free from artefacts good for a computer, but when watched on a TV, it does lack detail.

The best method I have found so far is using AVISynth and using a Spline36 resize after a smooth de-interlace. This is very near the end of sharp footage without artefacts, but looks great on a TV.

If your final footage is for a TV, don't judge the output by watching it on your LCD monitor is my advice as well.

Regards

Phil

[Chris Harding]

Thanks Phil

Ok at the moment I run all my AVCHD files thru NewBlue Upshift and save them as MPEG2 progressive files at 50mbps. I am therefore bringing in m2t files onto the timeline (My little DuoCore struggles with the MTS files from the camera)

So am I correct in saying that I need to do an intermediate render out to Sony YUV using the PAL Widescreen template and then imort each clip again into a new Vegas project???

When I render the second time do I simply render the Sony YUV file out to the usual MainConcept MPEG2 and again use the PAL DVD Widescreen template ???? (but enable reduce interlace flicker and increase the VBR to 9000) ????

So you say that using Vegas this way will still produce the end result as fuzzy rather than sharp????

I agree $4500 is a little harsh to have slightly better video..all my projects are weddings so the end product is always DVD. Admittedly I have never had a bride complain about quality and most own decent LCD or Plasma TV's at least 42" or larger too!!

I'll certainly try a test clip using the double render and see if the extra effort is worth it!!

Great advice

Chris

[Phil Lee]

Hi

The Mainconcept reference encoder uses the Mainconcept SDK, this SDK (for MPEG2/4) is the same as supplied with all software that uses Mainconcept encoders, that includes Sony MovieStudio for £40.00! Mainconcept's main revenue is from selling their SDKs and not the reference encoder, which is simply so over-priced it is untrue.

What the Reference encoder is doing is better resizing and removal of interlacing artefacts before it sends the footage to the encoder (although at a cost to detail), however if you feed the Reference encoder progressive footage directly, as will happen with your de-interlaced source, it will bypass any of the filtering for these artefacts as it just does a resize under the assumption it is true progressive footage, then extracts the fields, so it will probably look little different to Sony Vegas default output.

What I would do is give this a go:

1) Export your finished edit by Render As... then select Save as type as Windows AVI, select PAL DV Widescreen template, then click Custom.

2) Field order needs to be Top field first, pixel aspect should be 1.4568, Video format select Sony YUV Codec. Under the Project tab video rendering quality should be Best. Save it out.

3) Now bring this back to a new project, right click the now single clip and tick "reduce interlace flicker", then this time render out as MPEG2 DVD, check the Project is set to Best again, and under Video make sure a tick is in Prioritize quality over speed, Video Quality slider all the way to High, and I'd go for Variable bit rate, with a maximum of 9500 and average 9000.

The reason I would recommend doing it this way is the "reduce flicker" has little effect on HD footage, which I assume is because it applies the filter before the resize takes place, but the interlacing roughness is created after the resize when the SD footage is turned into interlaced. So by these separate steps the "reduce interlace flicker" is forced to act on the SD footage.

Whatever you do the end result is never as good as what you might expect or would see from commercial DVDs. This is because commercial DVDs are normally from progressive sources, and the initial capture can be made at the correct resolution from the film, not to mention they will encode scene by scene on very expensive equipment.

The problem is to get to DVD interlaced resolutions from HD interlaced, it has to be de-interlaced to 50fps, then resized to non-square pixels, then those 50fps are re-interlaced to 25i by taking odd lines from frame 1 and even lines from frame 2 and so on, but this isn't really interlace footage, by now its more pseudo interlaced. Those fields when turned into progressive again on our TV or computer screens just rattle around inside one another making it hard to be de-interlaced cleanly, so we either get very rough and artefact strewn footage, or these artefacts are essentially smudged out by removing detail giving cleaner footage without the artefacts but with less resolution.

You can see this by a quick test. Take some of your progressive 50fps footage to the time-line, right click on it and set the sample rate to 0.5, this gives you true 25fps and avoids Sony Vegas merging 2 frames to get 1 which doesn't look good! Then go straight to MPEG2 DVD but select progressive, then watch this back. It should look pretty good, except for stuttery movement due to the lower frame rate.

Regards

Phil