View Full Version : Need a way to to make AVS for 1080i -> DVD (MPEG2) conversion
matt163201
4th April 2009, 14:26
Hey guys I have a couple of HDTV captures and I want to make them standalone DVD compliant.
The source is 1080i, its 29.97 pic/s and 29.97 frames/s with a 2:3 pulldown
I want to reduce the resolution to 720x480 NTSC and put a Inverse Telecide AVS filter on it so that the pics/s plays at 23.976 while keeping the frames/s at 29.97
Perfect progressive playback, Im pretty new to AVS, tho Ive used it in GK with Xvids, pretty easy but for this I need to import it into VirtualDubMod and use either CCE or Procoder to do the compressing
Can anyone help me out, I know how to do filters for Xvid and x264 like I said but Im not a pro
Thanks for any help
neuron2
4th April 2009, 14:32
Do your captures have hard pulldown or soft pulldown?
matt163201
4th April 2009, 14:55
Im not 100% sure what you mean, so hopefully I can give you enough information to decipher the answer
ALL of them have 3 progressive frames followed by 2 interlaced frames.
In Gspot, some of them (most) follow this... the pic/s is 29.97, frames/s are 29.97 and fields/s are 59.94
I know I can force film in DG Index and get the ones with 29.97 pics/s down to 23.976, but from what I understand this should not be done and its wiser to use the IVTC script in AVS. Ive used Procoders built in GUI with preset filters IVTC but sometimes the results are not as good if I use AVS in GK if I make an Xvid instead, the problem seems to lie with the IVTC filter
This is the info I have from Gspot, when I play the unaltered capture M2V demuxed, I can see jaggies, im sure once I use the IVTC and then make it into DVD, everything will be fine.
This is what Im hoping to achieve
My DVDs VOBs/ M2V file come up as 23.976 pics/s, 29.97 frames/s and 59.94 fields a second, I believe this to be the best output for making my DVDs, I know not all NTSC DVDs are like this but I would like to make all mine like this, as Im fairly confident this is the correct settings, since this is what looks best when I watch it, even with the crappy IVTC in Procoders GUI, if I output them with 29.97 pics/s then I see jaggies, and I know IVTC is the way to go, no other de-interlace method
neuron2
4th April 2009, 15:56
Find out what soft versus hard means and then answer again.
Or upload an unprocessed source sample.
matt163201
4th April 2009, 19:24
Thanks for your help
Im gonna take 1 minute sample clips and upload them and put the links in here, theyll be in m2v form straight from DG Index with honor pulldown checked, that way you can see some of them Im dealing with
Thanks, Ill post them when I get home
neuron2
4th April 2009, 19:26
When you demux to M2V with DGIndex, the field operation is irrelevant because the stream is demuxed as is.
Do a preview in DGIndex with Honor Pulldown Flags. If the field repeats box counts up then you have soft pulldown.
matt163201
5th April 2009, 04:58
Hey, I checked out the demuxed files in gspot and I was wrong
They all report
23.976 pics/s, 29.97 frames/s, and 59.94 fields/s, 3:2 pulldown, and interlaced
All of them are top frame first except one which is labeled as such and was also taken from a different channel than the rest of them, files are labeled as such, all are m2v files cut with DG Index, all around 100MBs a piece
http://www.megaupload.com/?d=HMJW33YD
http://www.megaupload.com/?d=4111NGMX
http://www.megaupload.com/?d=4BQ0DMXA
http://www.megaupload.com/?d=CQBYPUPB
All samples are approx 1 min, hopefully this should give you enough information to help me.
Also is there any step guide to go from these 1080i's to DVDs? using Procoders encoder and what app to use to do so (such as VDubMod)?
Thanks
Alex_ander
5th April 2009, 08:48
...is there any step guide to go from these 1080i's to DVDs?
Your first sample is basically soft-telecined film (=encoded as 23.976 with field repeat flags added to display it as 29.97) but probably with some edits applied after file creation. So at preview DGIndex only shows 82% film by the end of the cut. If it were closer to 100% it would be more safe to apply 'Force Film' mode for getting 23.976 directly into script (with those 82% some frames will have traces of interlacing). Then better use 'honor pulldown flags' in field operation settings in video menu (same mode as used for hard-encoded telecine).
So your steps with this particular file are:
1.Load the mpeg into DGIndex. Set 'honor pulldown flags' (probably it is default) in video/field operation, set 'demux all tracks' for audio. File -> save project. You'll get a d2v project file + demuxed audio.
2.Put DGDecode.dll (same as used with DGIndex) and ColorMatrix.dll (external plugin) + Decomb.dll into plugins folder of AviSynth installation directory. Open the created d2v with script*:
MPEG2Source("[path]\sample a (TFF ch. a).d2v")#29.97
Telecide().decimate(5)#23.976 from here on
LanczosResize(720,480)
ColorMatrix(mode="Rec.709->Rec.601")#adjustment for HD->DVD
ConvertToYUY2()#for CCE or Canopus
* to make things easier you can install AVSP - excellent free script editor with video preview, then simply drag d2v file onto its window to have first line of the script line written (no need to set path to the file). Then add the rest of the lines.
3.After testing the script in VDub or AVSP encode video as 23.976 elementary stream with 3:2 pulldown setting (same soft pulldown as in source file). E.g. in CCE (I don't use Canopus, be sure to turn off blending mode there) the settings would be 23.976 for framerate + 'For DVD' (turns on 3:2 pulldown), max GOP length=12. Take audio delay value from demuxed file name and use both streams for authoring.
matt163201
5th April 2009, 11:32
Hey thanks!
I think I got what I need the ONLY other thing I need to add in I think would be crop but Im gonna get that app you said to get, I also got TMPEGEnc 30 day trail just in case I needed it
I was looking in the CCE teplate, should I be using NTSC DVD or NTSC DVD w IVTC? Also I did not see any option that said "For DVD" but I did see the option for 2:3 AND 3:2 pulldown
Im using CCE 2.70.02.12
Edit: Also why would I need to do color correct when (according to DG Index) the color is ITU-R BT.470-2? I thought this was a form of 601 color and not 709?
NerdWithNoLife
6th April 2009, 03:17
Are there any commercial cuts/glitches in the stream, because this could cause audio sync issues if not handled correctly.
matt163201
6th April 2009, 03:36
No, no commercials, no cuts.....its a straight shot, I only trimmed alittle off the beginning and end thru DG Index, I dont have to cut anything else
NerdWithNoLife
6th April 2009, 06:06
Also know you can use HCEnc, an excellent program by hank315 which you can find in this thread. (http://forum.doom9.org/showthread.php?t=137666)
neuron2
6th April 2009, 14:37
Force Film mode is an option here too because the non-pulled down sections are short and low-motion.
matt163201
6th April 2009, 15:07
@neuron
These are just samples, they are movies, so guessing the length of them is around or over 90 mins, and Ive checked all of them and NONE of them report over 90% film in DG index, the closest one gets is around 80 so force film in DG is not an option, can you give alittle more insight perhaps....or do you agree with what was mentioned above?
neuron2
6th April 2009, 15:27
If you are going to just dismiss what I say, don't bother asking me anything.
matt163201
6th April 2009, 15:53
If you are going to just dismiss what I say, don't bother asking me anything.
neuron, I wasnt dismissing what you said, I think I misunderstood what you said, when you said the samples are short, I thought you meant in length, thats why I was telling you the the runtimes, sorry if I was mistaken
Could you explain further? So should I force film in DG and not use Avisynth? Also I just took a minute clip, there may be fast paced scenes in the movies, if that helps any
Sorry again if I misunderstood
neuron2
6th April 2009, 16:33
I meant that the non-3:2 sections in the first sample are short sections where there is little motion. So using force film will not degrade them significantly. You can try force film and see if you like the results.
Anyway, using an external IVTC is going to decimate the same way, so if force film does not let combed frames through, then there is no reason not to use it.
Alex_ander
8th April 2009, 21:49
I was looking in the CCE teplate, should I be using NTSC DVD or NTSC DVD w IVTC? Also I did not see any option that said "For DVD" but I did see the option for 2:3 AND 3:2 pulldown
Im using CCE 2.70.02.12
No, the template with IVTC is for encoding directly from a hard telecined source (not your case) by previously removing pulldown.
For soft telecine target (with source video already IVTC'ed) it is better to begin with PAL template (since GOP size is 12 there by default, which is recommended for soft telecine), then manually set framerate 23.976 + 720x480 or use 'For DVD' checkbox if available (I have it in version SP 2.70.02.04, not sure it is there in Basic version). Try to check 3:2. If for some reason flags are missing in the output 23.976 video (GSpot will show), it will be possible to quickly add them in DGPulldown utility (also by neuron2).
Also why would I need to do color correct when (according to DG Index) the color is ITU-R BT.470-2? I thought this was a form of 601 color and not 709?
Then of course don't use ColorMatrix.
matt163201
9th April 2009, 16:23
Alex,
Im using CCE, got my script done I created a custom template...MPEG2 for DVD, FR @ 23.976 Size at what I need Pulldown @ 3:2 with Inv 3:2 checked and greyed out, but I couldnt find GOP size, is it SEQ header under Advanced in how many GOPs? Been reading the guide here on doom and the user manual for CCE....is there anymore settings I should be looking at under advanced and picture settings? Also GOP sequence @ m=3 and nm=5
Im trying this both ways to see which gives the best results
Alex_ander
9th April 2009, 16:48
You need m=3, n/m=4 in advanced (this also means GOP length 12). If you have "For DVD" checked and 23.976 set, CCE should apply correct pulldown without other manual settings.
matt163201
9th April 2009, 17:14
Alex,
Thanks, Im pretty sure I have all the settings right then, did you get a chance to look at the other samples? most notably 3 (which is much different from the rest) and 4 (which is from a different channel and i think is bff)
Also I was looking here and it says m=3 and nm=5 for NTSC, for PAL it would be m=3 and nm=4
http://www.doom9.org/index.html?/mpg/cce270.htm
Which should I do? 4 or 5?
Alex_ander
9th April 2009, 19:14
I had a quick glance at 'C' clip. The difference from the rest is that it's hard-telecined (encoded extra fields). This means that 'honor pulldown flags' is the only choice for working with it in DGIndex (however I'd still prefer this option for the other clips for safety).
Also I was looking here and it says m=3 and nm=5 for NTSC, for PAL it would be m=3 and nm=4
These are default settings in NTSC/PAL templates (15 and 12, maximum possible are 18 and 15) used for 29.97 and 25. As for encoding 23.976 with pulldown flags, lower number of 12 is recommended (in CCE same as default PAL setting, 4 instead of 5). The reason is explained in GOPNote.txt file included with DGPulldown docs.
matt163201
9th April 2009, 20:23
Alex,
Thanks a bunch for your help
I think I now have enough info to FINALLY do an encode in CCE with my AVS and have all the correct options set. I knew sample 3 was gonna be different cause it only reported like 20% film and I have a feeling the source was a LD.
Im gonna try both methods of Force Film (on the ones that apply, not sample 3) and letting AVS do the pulldown for me and run them thru CCE multipass VBR and see which delivers the better pulldown (since the quality will be about the same, just looking for interlaced jaggies)
So again, I would like to thank you Alex and neuron for all your help and guidance, and Ill probably end up taking information from both of you to implement on these transfers and ill report back
vBulletin® v3.8.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.