CCE 2.66 and Pulldown Detection

[wmansir]

Has anyone tried the new pulldown detection on Hybrid material? If it works the way I think it does, it may be the best solution (IVTCing the FILM and leaving the Video alone). I'm going to start working on the Larry Sanders show DVD soon, so I'll know myself, but I would like to know if I'm correct in my thinking.

Also, has anyone gotten pulldown detection to work with multi-pass? I keep getting an error about "the pulldown does not match the previous pass" after the .vaf creation, just as the first real pass starts. I tried it with a couple of simple .avs scripts and one huffyuv .avi. I'll have to experiment more.

Thanks,
wes

(This is an almost copy of a post from the main 2.66 thread, but I wanted to focus more on this promising new feature of 2.66 so I started a new thread.)

[wmansir]

I have experimented a little and this feature is very exciting. I have yet to try it on hybrid material so I don't know how it will affect the video portion of the stream, but on telecined source it works excellent. However, I must admit most of my sources are film based, which is easy IVTC. I'll be interested to see how it handles the more complicated anime that torture decomb and greedyHMA.


I found a section of the .pdf which describes the error I have encountered with multipass and pulldown detection. On page 42 it states:
When encoding in multipass VBR mode, make sure to recreate
video information file if you manually changed inverse
3:2 pulldown setting. If an existing video information file is
used, an error message like Fig. 3.19 is shown.
Fig. 9.19 is exactly what I am getting, however I haven't changed any settings. This error is probably related to the fact that the pulldown detection menu has not been generating a pulldown list. The list is aways blank when I open the Pulldown detection settings menu, even after a .vaf has been created.

Am I the only one encountering this problem? Please let me know one way or the other.

[wmansir]

I just tried a couple of things to see if having an 'altered executable' has any effect on my Pulldown detection + multipass problem. Here's what I learned:

A clean install of the trial program will generate a pulldown list when using CBR encoding, that's the good news. The bad news is it does a horrible job at detecting pulldown, probably due to the fact that the CCE banner it sticks onto the video is messing up it's pattern detection. This makes the trial program virtually worthless, since Pulldown detection is the biggest selling point and the trial program can't even do it correctly.

More bad news is that even though the clean install generates a pulldown list with CBR, multipass still fails. I'm getting the same error and weather I use the .vaf from the CBR pass or create a fresh one, it doesn't work. Also, the .vaf creation pass doesn't seem to create a pulldown list at all.

After a great sea wave produced by submarine earth movement altered the executible it doesn't create a pulldown list in either CBR or .vaf creation, however Pulldown detection works properly.

Well, that's about all I know right now. Feel free to add any info, I need some sleep.

EDIT: Just so it doesn't look like the whole point of this thread is to inflate my post count I'm going to just edit in these comments...

While I was sleeping I encoded one pass vbr with Pulldown detecion (an episode from buffy season 2 dvd) and to my surprise the file was exactly the right size. Unfortunatly, muxing with BBmpeg with the standard SVCD settings resulted in audio sync problems. It starts out fine, but looses sync quickly. However, TMPG muxes the stream fine. For some reason it's also 5MB smaller than the BBmpeg output.

[Vapor]

2.66.01.02 Released Today.

From the readme:
<< Known bugs >>
1. VBR encoding with "3:2 pulldown detection" selected causes error.

[Gpz1100]

Vapor,

Where does one get the update? On their website they still show 2.64.xx

[wmansir]

Go to the Japanese verison of the site. Half the site is still in english so you can find the download section easy. Or you can try this direct link to the trial: http://www.cinemacraft.com/files/cce/ccspt266.zip

[mgd72]

Hi,

Could someone please give me a quick guide on how to do the Inverse 3:2 pulldown in CCE? I left everything at default and checked 3:2 pulldown detection but it isnt working for me. The output looked terrible. Everything had the interlacing artifacts in it. I even left the settings it detected on the first encode of the CBR and encoded it again, but no luck. Im totally lost :) Im trying to encode Star Trek: The Next Generation episodes which are hybrid.

My steps:
1) Use DVD2AVI with Force Film OFF
2) Use VFAPI to make the avi file
3) Load avi into CCE and set to CBR 2200 w/ 3:2 pulldown checked

Also, If avisynth will work with the new CCE, I would like to try that and use it to Bicubic Resize to SVCD before going into CCE, but not sure if that would screw up the pulldown detection or not.
Any help would be greatly appreciated. Thanks

Mgd72

[Vapor]

Have you tried "decomb" for your IVTC?

AviSynth 2.06 does work with 2.66.01.02 (others versions may also but I've only tested 2.06).

Resizing frames before IVTC will knacker the detection (does for me but I prefer decomb anyways) but a resident guru may confirm or deny this.

[wmansir]

I got it working, but it didn't work with the unpatched trial version (2.66.01, haven't tried .02). Also, I only resized horizontally (720x480->480x480), which should be safe. Vertical resizing blends the interlace lines and makes IVTC impossible.

[mgd72]

well Im trying not to use decomb. I have tested that filter 30 different ways and it still messes up the movie because the space shots are NTSC and the character shots are FILM. So one or the other is always jerky when using decomb.

Credit goes to daxab for posting a way to get these types of shows to work great in Tmpeg, but I am trying to find a CCE alternative so I can get better quality. here is a post where the problem has finally been solved using Tmpeg:
http://forum.doom9.org/showthread.php?s=&threadid=34924&perpage=10&pagenumber=2

So the solution is using Inverse 3:2 pulldown and keeping the shows at 29.97fps. Now if CCE can do the inverse pulldown, then I will be a happy camper :) The problem is its not working for me..

[wmansir]

I understand what your trying to do, and the new pulldown detection should do this. The first step for me was to patch the trail, because as it was it did not detect the pulldown properly. In fact, when I looked at the pulldown list, under the pulldown detection settings, it though most of my source was video, even though I knew it was FILM.

After reading the thread you linked I think you already know this, but the NTSC video portion of the source will still have interlaced lines, only the telecined film portion will be restored to progressive frames. I can't really offer much help if it isn't working (other than to patch the trial) because I too left the settings at their defaults.

[mgd72]

ok thanks. I tried patching it just to see and still no luck. It still shows the interlacing artifacts. Maybe I am thinking this is something different than the Inverse 3:2 pulldown setting in Tmpeg. It just done a terrible job on the episode I tried. Oh well back to the drawing board. Thanks for trying :)

Mdg72

[Vapor]

If you've found a way in Tmpgenc to sort IVTC but want to encode with CCE try getting the job ready in Tmpg (the IVTC bit) save a project (.tdr) then make a pseudo avi of the .tdr using vfapi. Drop the avi into CCE and encode (or frameserve again using avisynth - this will be slower still though).

IIRC I used to do this before well developed IVTC plugins for avisynth came out and it worked ok. I believe the .tdr maintains the IVTC info needed.

Might not work but worth a try.

[jdobbs]

@mgd72

If you've found a way to make TMPGENC work with StartrekNG I'd sure like to here the details. After reading the link I tried about every combination of settings I could imagine and it seems to act the same way as my attempts with decomb.

1. Are you feeding it directly from the M2V or are you using DVD2AVI to frame serve?

2. Are you setting FORCE FILM or NONE if using DVD2AVI?

3. Output setting set to 29.97?

Anything you could give me would be appreciated. TMPGENC doesn't give the same quality as CCE, but to me a jumpy picture is a lot more annoying than a few artifacts.

Thanks,
jdobbs

[mgd72]

Hello,

Here are the steps I used:

1) Smartripper to rip each episode to hard drive.
2) DVD2AVI with Force Film OFF
3) Opened d2v file into Tmpeg then changed all the settings.

I have a Tmpeg template saved for these if you want to try it. Its for a SCVD of 1 episode per CD. I cut the beginning and end credits out also to get a little more bitrate. But in a nutshell, I used the SVCD settings with a couple minor changes:

Video Tab:
Encode Mode - set to Inverse 3:2 pulldown
DC Component precision - set to 10 bits
Motion Search precision - set to High Quality
EDIT: framerate should be set to 29.97

Then under the advanced tab, you have to make sure it sets those right. I have to manually change mine on some of it. For the ST episodes they should be:
Video Source Type: Interlace
Field Order: Top Field First (field A)
Source Aspect Ratio: 4:3 525 Line (NTSC)
Then check the source range box and cut out the credits if you desire.

AUdio Tab:
no changes, but I use 192kbs.

All other settings are SVCD default. One thing else to double check is the vbv Buffer Rate on the video tab. I set mine to 112.

I think this is everything. I just want to mention that you might want to burn a test clip to CD and play it on a standalone player. I noticed some jumpiness when playing it on my computer, but it played fine on the TV. Could just be my computer though...

Hope this helps. let me know if you need the template and I can email or icq it to you. Or post it here if the mods will allow it.

Mgd72