Best way to get .ts and .tp into divx or Xvid?

[Zep]

Anyone know the best (or darn good hahah) way to get.ts and .tp into divx or Xvid?

I just bought a MyHD 120 HDTV PCI card and I'm having a hell of time geting the streams into a format i can load into virtual dub for editing and then encoding into Divx or Xvid.

Does anyone know a simple way and could you list the steps and apps you use?

A BIG THANKS and I mean it cause my head is hitting the monitor on this one and it hurts :)


Thanks,


Zep

[timecop]

http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=227837

Just ignore all the WMV9 related junk, and read what to do with .ts files.

Short summary, load .ts into DVD2AVIT3, save project/demux audio, use avisynth to load d2v, encode.

[holden]

I Like Jonny's Enc 1.03 to save on repetitive keypresses, even
more than just Vdub batch processing. Note that scripts are
read once at the beginning of a pass, thus you can change,
for example, the file reference, such that one mouse click
away is processing for that file. Whammo!

Anyone else got any speedup suggestions?

HF

[Zep]

Originally posted by timecop
http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=227837

Just ignore all the WMV9 related junk, and read what to do with .ts files.

Short summary, load .ts into DVD2AVIT3, save project/demux audio, use avisynth to load d2v, encode.

yes that is pretty much how I ended up going.

DVD2AVIT3 and make a d2v

avisynth to frame serve to VD/VDM

then Xvid or Divx encode

My big complaint is that resize is so slow and so is Telecide
on 1900 x 1080i stuff that frames serving even on a 2.4Ghz P4
with a gig of ram is like 8 FPS.

now for 720p it flies cause the one step is removed and the resize
is on a lot less pixels. I wish everyone was 720p. :)

[trbarry]

If you are mostly concerned with speed and can live with a 960x544 output resolutiion then get rid of the Telecide & Resize, substituting.

YV12InterlacedReduceBy2() # reduce to progressive 960x544 (from 1080i)
# Decimate(cycle=5,quality=0) # optional, 24 FPS telecined material only

Then crop or use FillMargins(x,x,x,x) if there is garbage on the edges.

- Tom

[holden]

I can tell you ZEP and TRBARRY don't have 1366x768, or you'd never say that. The increased quality of that resolution over 720p or, God forbid, 544 is nearly as appreciable and large as moving from SD to HD. It maybe a noisy box, but I love my Sony HS-10. Pretty inexpensive, too.

I maintain 768p-processed stuff (generally at 6Mbps, Xvid/DivX) looks better than the 1080i original in many cases (though, of course, not all of them).

HF

[Zep]

Originally posted by trbarry
If you are mostly concerned with speed and can live with a 960x544 output resolutiion then get rid of the Telecide & Resize, substituting.

YV12InterlacedReduceBy2() # reduce to progressive 960x544 (from 1080i)
# Decimate(cycle=5,quality=0) # optional, 24 FPS telecined material only

Then crop or use FillMargins(x,x,x,x) if there is garbage on the edges.

- Tom

thanks i will give that a try :)

[Zep]

Originally posted by holden
I can tell you ZEP and TRBARRY don't have 1366x768, or you'd never say that. The increased quality of that resolution over 720p or, God forbid, 544 is nearly as appreciable and large as moving from SD to HD. It maybe a noisy box, but I love my Sony HS-10. Pretty inexpensive, too.

I maintain 768p-processed stuff (generally at 6Mbps, Xvid/DivX) looks better than the 1080i original in many cases (though, of course, not all of them).

HF


You missed the point I was making where as trbarry understood the
point I was making. I do NOT want a higher rez. I said how slow it is to resize
down 1920 x 1080i and that i like 1280 x 720p because it is faster to resize.

here read the part again

"720p it flies cause the one step is removed and the resize is on a lot less pixels"

see where I said resize? i am going even lower than 1280 x 720p and why his
suggestion is a one i am going to try.

resizing 1360x768p down would be SLOWER to resize than 1280 x 720p
so i do not want that rez. it is much better to choose a rez that is closet to the
final rez i want since it is faster to work on less data and you get less blurring
on the scaling etc...

[timecop]

Originally posted by Zep
My big complaint is that resize is so slow and so is Telecide
on 1900 x 1080i stuff that frames serving even on a 2.4Ghz P4
with a gig of ram is like 8 FPS.


That sounds about right. on my 2.6ghz p4, it takes about 40 minutes for 25 minutes of video, which is ~15fps or so (this is assuming resizing down to SD resolutions of 640x480 or so). For QHD (is that what we should call 960x544?), its about an hour for 25 minutes.

On upscaled SD, separating fields, then selecting every other frame, resizing and then decimating (if neceesary for film), seems to be a very fast way to get quality looking contents.

[IWatchHDTV]

Zep, I'm curious what settings you're using. Are you doing something like Divx 5.1 standard/slow, 2 pass encoding, with light PV, GMC + bidi? I'm getting decent fps on standard pass, but on the slow pass, I'm barely getting 1 fps :mad: With Divx 5.1.1, it's a little better, I'm getting around 2 fps :p I want to get some more numbers on how other ppl are doing with Divx 5.1 and HDTV.

[jrmann1999]

I can tell you after extensive testing there are two ways(possibly lossy, but never in my case) to very quickly get 1080i material into mor manageable formats.

#1 - Use the new Uncomb filter. It will significantly reduce the time compared to using telecide to pull your material out of interlaced and prep for a decomb.

mpeg2source(bleh)
uncomb()
decimate(cycle=5)

will pull 29.97 1080i into 23.976 1080p material.

#2 - This method is a bit more risky as you are blindly going to do some field work(not that uncomb doesn't as well) and you might get a few interlaced slip throughs, but in the 8 months I've been using it I haven't seen 1(and I get 2-3fps more than using uncomb)

mpeg2source(bleh)
assumefieldbased()
separatefields() #pull 29.97i to 59.94p
selectodd() #because above we effectively doubled fields - 29.97p
decimate(cycle=5) #pull to 23.976
assumeframebased()

what you are left with is a 1920x540p feed that you can easily and quickly manipulate into whatever smaller resolution you like. If you are going to xvid or DVD the 1920x540 is plenty to resize with imho.

[abman]

I use that method # 2 and have had no significant problems with it. (thought I use selecteven() instead of selectodd(). don't think that matters). No noticable interlacing is left in the final resized video. never messed with uncomb, but the separatefields method is way faster than telecide.
You can also get a boost by not putting your avs directly into xvid or divx. I've noticed going to avs > huffey > xvid or cce is faster than doing an avs straight to xvid or cce.

[jrmann1999]

HOw is that a speed boost when it'll take you awhile to encode into huffy, unless you're talking 5-10m for ts->huffy, and what sort of speed are you getting from huffy->(divx,mpeg) ?

[abman]

I guess its just faster not having to do the avs process and the xvid bit allocation at the same time. I don't have the FPS numbers on me to back it up, but I shave a good 20-25 minutes off the total time for a 45 min show encode. It probably also matters how much is in your avs. if you're doing a ton of processing, such as telecide, decimate and a resize then you'll gain more this way.

[trbarry]

mpeg2source(bleh)
assumefieldbased()
separatefields() #pull 29.97i to 59.94p
selectodd() #because above we effectively doubled fields - 29.97p
decimate(cycle=5) #pull to 23.976
assumeframebased()

what you are left with is a 1920x540p feed that you can easily and quickly manipulate into whatever smaller resolution you like. If you are going to xvid or DVD the 1920x540 is plenty to resize with imho.

If you are working with YV12 data then just separating the fields can cause a mild chroma delay. YV12InterlacedReduceBy2() will correct that. But for all I know maybe Avisynth's SeparateFields() also does that now.

If you really want the full 1920 width then the YV12InterlacedReduceBy2 module also contains a YV12InterlacedSelectTopFields() function. This will take only the top fields (leaving 1920x544 @ 30 fps) but also adjust the vertical chroma delay. Then you can still do the horizontal resize yourself when you want something other than 960.

- Tom