BEST way to convert HD video to standard DVD?

[edDV]

1280x720p video is broadcast at 59.94 fps natively. 704x480p video is also broadcast at either 59.94 (live) or 23.976 (film*) natively for ATSC.

DVD only supports 480i 29.97 or 480p 23.976 (optionally telecined to 29.976 480i). BlueRay can handle 480p or 720p at 59.94.

So, to encode for DVD, you need to convert progressive 59.94 fps to interlace 29.97 fps. Each field should be pulled from separate frames for smooth motion.


* Although 23.976 progressive (1080p, 720p, 480p) is an approved ATSC format, nobody is broadcasting it as yet. Two 720p 23.976 fps movies could be broadcast on one 19 Mb/s ATSC channel in high quality.

[senorroboto2k5]

Thanks for the info. I tested the SelectEven script and it worked well 59.94->29.97.

I just edited my previous post with newer pics from using DVAVCDec/HCenc, please check it out :/

[Soopafresh]

Your colors are changing, and the reasons could be several. The easy (but imprecise) workaround is to add this to your script:

tweak(sat=1.1,cont=0.9)

[senorroboto2k5]

Thanks. That brings out the color to around the source, but it's still very blurry.

LimitedSharpen didn't really do much either in this department.

I would be OK with this but the transition is nowhere near as good as FulciLives, so I'll keep on trying.

[FulciLives]

Thanks. That brings out the color to around the source, but it's still very blurry.

LimitedSharpen didn't really do much either in this department.

I would be OK with this but the transition is nowhere near as good as FulciLives, so I'll keep on trying.

It might just be the way you are capturing the images.

In my case I use VirtualDubMod as it has a feature that allows you to save a frame directly to a PNG image file. It can natively load the m2v file that HCenc spits out so getting a snap shot of that is easy. To get a snap shot of the source I use my AviSynth script (the one I use to encode the source with HCenc) but I delete any filtering or resizing so I get a "raw" unprocessed image from the source.

When taking a snap shot from the m2v file try and do it from an "I" frame aka a key frame. The "P" and "B" frames will never be as sharp as an "I" frame. That is just how MPEG-2 works.

- John "FulciLives" Coleman

[manono]

Your colors are changing, and the reasons could be several.

I take it he's not taking into account the colorimetry change from Hi-Def BT.709 to Standard-Def BT.601 when doing the conversion. That would be one explanation for the slight color changes

[jagabo]

5000 kbps might be too low for a basketball game.

Fixing your colors with Tweak() is a kludge. You should just get the colorimetry right in the first place. Manono is correct: you need to convert rec.709 to rec-601 with ColorMatrix(mode="Rec.709->Rec.601").

The blurriness has little to do with the resizing filter. It is the way you are viewing the videos and the way you are creating snapshots. Keep in mind you are reducing the frame size from 1280x720 to 720x480. You are losing nearly half your resolution. Of course the smaller resolution will be much fuzzier when viewed full screen. You need to stop posting snapshots that are resized from the videos' native resolution. Use a program that gives you one-to-one pixel mapping in the snapshot. Post a 1280x720 image of the source, a 720x480 image from the resulting MPG file. 800x500 screen dumps from WMP are extremely unhelpful

Even better, post a few seconds of the HD source and someone will give you an AVS script that handles the video properly. Then if you have questions about the script they can be answered. As it is we're all having to guess at what you are doing.

[senorroboto2k5]

OK sorry I've been ignoring the "right" way to capture the images...

Source


DVD


^They're being resized

It does look pretty fine. Almost dead on.

Still, there's a difference in watching them on WMP and VLC. Is it because it alternates between I frames and non-I frames? Or is it just because there's something wrong with comparing the full-screen source and the full-screen MPEG2?

Anyway I'll try to cut out a few seconds and post it on here. Does it matter where I upload it? Sendspace or rapidshare OK?

[senorroboto2k5]

Here's a 5-second or so clip. It captures the game and an up-close shot of the player.



Let me know if there's another way to upload it



4_797_804.mp4

[jagabo]

OK sorry I've been ignoring the "right" way to capture the images...

Source...
DVD...

Ah, much better. As you can now see the MPEG image is pretty sharp in its native form. You may get slightly sharper or slightly fuzzier results depending on what resizing filter is used. TMPGEnc's bilinear is one of the poorest. AviSynth's LanczosResize() is pretty sharp without introducing moire artifacts. Etc.

However, in motion, the MPEG2 STILL looks blurry. The overall difference that exists still lies in the previous pics that I posted, i.e. watching the source is like looking at that pic, and watching the converted MPEG2 actually looks like the blurry second pic. Is this a problem, or is this the way that it's supposed to be?

A few things may be going on. When you indicate the MPEG2 looks blurry when watching -- are you watching full screen? Yes it will look blurrier than the HD source if enlarged. You may have other problems which make it worse. The player may be deinterlacing while it plays. The player or graphics card may be using a poor resizing filter.

Are all the frames in the source that clear? The recent Olympics videos were somewhat unusual in that they all came from China as 50 Hz HD. When broadcast in the USA they had to go through a frame rate conversion to 60 Hz. The technique used generated fuzzy double exposure-like frames when there was lots of motion. This should be visible in the HD video at times (the SD MPEG shouldn't be any worse).

Anyway I'll try to cut out a few seconds and post it on here. Does it matter where I upload it? Sendspace or rapidshare OK?

I like Sendspace.

Also what's the best program to cut out a few seconds of the mp4/h264 file?

AviDemux should be able to cut out a small piece. Open the video, use the slider and the mark-in and mark-out buttons to mark a section, set the output to MP4, save.

[senorroboto2k5]

Here's a Sendspace link.

http://www.sendspace.com/file/hp3w8v

I think I also attached it to the post above. 5.something MB's

[jagabo]

Yes, I noticed the upload in the earlier link. Downloaded it and looking at it now.

Here's a conversion to MPEG 2 with HCEnc:

sd.zip

I used the following simple script:

AvcSource("temp.dga")
ColorMatrix(mode="Rec.709->Rec.601")
LanczosResize(720, 480)
SelectEven()

That probably looks similar to your MPEG 2 encode.

If you want smoother motion you can use:

AvcSource("temp.dga")
ColorMatrix(mode="Rec.709->Rec.601")
LanczosResize(720, 480)
AssumeTFF()
SeparateFields()
SelectEvery(4, 0, 3)
Weave()

Encode that as 29.97 fps interlaced, top field first.

[senorroboto2k5]

Awesome. Thanks for the help!

[halsboss]

AvcSource("temp.dga")
ColorMatrix(mode="Rec.709->Rec.601")
LanczosResize(720, 480)
SelectEven()

AvcSource("temp.dga")
ColorMatrix(mode="Rec.709->Rec.601")
LanczosResize(720, 480)
AssumeTFF()
SeparateFields()
SelectEvery(4, 0, 3)
Weave()

They're both for a Progressive input ONLY, right ?? Otherwise you have to other things, like this PAL 1080i->576i (->DVD) TFF example (if you don't want to go the quality deinterlaced resizing route) :-

Global NewHeight=576 
Global NewWidth=704 
SeparateFields()  
Shift = (GetParity() ? -0.25 : 0.25) * (Height()/Float(NewHeight/2)-1.0) 
E  = SelectEven().Spline36resize(NewWidth, NewHeight/2, 0,    Shift) 
O  = SelectOdd( ).Spline36resize(NewWidth, NewHeight/2, 0,   -Shift) 
Ec = SelectEven().Spline36Resize(NewWidth, NewHeight/2, 0,  2*Shift) 
Oc = SelectOdd( ).Spline36Resize(NewWidth, NewHeight/2, 0, -2*shift) 
#E  = SelectEven().Lanczos4resize(NewWidth, NewHeight/2, 0,    Shift) 
#O  = SelectOdd( ).Lanczos4resize(NewWidth, NewHeight/2, 0,   -Shift) 
#Ec = SelectEven().Lanczos4resize(NewWidth, NewHeight/2, 0,  2*Shift) 
#Oc = SelectOdd( ).Lanczos4resize(NewWidth, NewHeight/2, 0, -2*shift) 
Interleave(E, O) 
IsYV12() ? MergeChroma(Interleave(Ec, Oc)) : Last 
Weave()

PS the biggest bitrate you can manage is best, if you want quality sports action (eg circa 9200k) (DVDs are cheap).