DC10+AVS+CCE = "Wavey" Output

[bewley]

I've captured a video clip from with my DC10+ and created a script that simply loads the AVI, resizes it, and adds borders. I load this script into VDUB, apply the Flip Vertically filter and set my start and end points. I then frameserve this.

I open the VDR file in CCE and use the following setting:

MPEG2
3 Pass
2600 - 2617 - 1246 (avg-max-min)

Under the advanced tabs I have:
Upper Field First
Image Quality Priority = 10

After the encode, I multiplex using bbMPEG with my 128 kbit MPA file and burn with Nero.

The video looks excellent, except where there is movement. There is this wierd wavey thing going on ONLY where there is motion. If rest of the picture looks excellent. I also noticed that scene changes have a sort of combing effect -- except that the combs are like thick black bands. I have seen jerkiness from incorrect field order and combing from bad interlacing -- but never this.

I tried opening the VDR in TMPGEnc and didn't notice the problem until tried to Deinterlace the file as a test. Everything except for "Double (adaptive)" produces the weird snake-like wavey output.

Has anyone seen this behavior and know of a fix?

Thanks!

[Khizhiy]

I think the answer is in how you resize your video. Mind that it is interlaced with your DC10+. You can't resize it without doing deinterlacing first.

[bewley]

Thanks Khizhiy,

Yes. That was my problem! Also, I found that MPEG2DEC.DLL has a FlipVertical routine -- no more VDub! Everything is perfect now.

Thanks!

[Bach]

Hi bewley,
Could you please give us a quick round up of the steps you are doing?
I'm using DC10 as well, so would it be possible for you to add the release of software you are using as well? Studio1 or 7? etc.

[bewley]

Bach,

Sure! I hope that you have some understanding of CCE & Avisynth, but if not, I'll be happy to answer any questions -- to the best of my meger ability.

First, I capture using Pinnacle Studio Version 7 (7.07.1 I believe) using maxed out settings: 640 x 480 NTSC, 6000 kbps video rate, CD Quality Audio, etc. On my little PIII-550 (256 MB RAM, 20 GB ATA/66 HD), I am dropping about 1 frame every two minutes. Once captured, I end up with several AVI files as such:

Video 1.AVI
Video 1_001.AVI
Video 1_002.AVI
(and so on)

From Windows Explorer, I rename these to:

Video1.00.AVI
Video1.01.AVI
Video1.02.AVI
(and so on)

Next, I load an AVS Script that was originally created using the FitCD program. Keep in mind that this script requires the use of the MPEG2DEC.DLL plugin.

LoadPlugin( "<drive>:<path>\MPEG2DEC.dll" )
SegmentedAVISource( "<drive>:<path>\Video1.avi" )
AssumeFrameBased.SeparateFields
TemporalSmoother(2,1)
BicubicResize(320,207,0,0.6,16,8,608,224).Weave
AddBorders(16,33,16,33)
FlipVertical
ConvertToYUY2
Trim(143,20632)

NOTES:
* I use SegmentedAVISource, that is why I have to rename the segments after capturiung them.
* I am not sure if I should use the TemportalSmoother statement before or after the SeparateFields statement... It is my understanding that doing it after the resize will make the script run faster, but can also "damage" the output and make it look less sharp..?
* The SeparateFields and Weave statements were the pieces I was missing when I started this thread.
* I am resizing to 320 x 414 and adding boarders to make it 1/2 DVD NTSC size (352 x 480). This produces slight black letterboxing borders on the top and bottom of the video, but the bars on the left and right sides are "hidden" inside the TV (at least on MY TV). Note: this resize takes into account the way that CCE shifts the input up one line -- 352 x 416 is more "proper" and the resolution I would use if serving to TMPGEnc.
* FlipVertical, well, flips the output. If you have ever loaded a DC10+ capture through Avisynth, you will know that it serves upside down!
* You do not NEED the ConvertToYUY2 statement, but you do NEED the ConvertToRGB statement when serving to TMPGEnc.
* I left the Trim in -- I was doing an 11 minute clip from my last capture and this is how you set the start and stop frame.

Now, open this script with CCE 2.5 -- you will need to tunnel through AFAPI if you are using a version of CCE higher than 2.5 -- and fill in your settings. I always disable the audio, use MPEG2 VBR 3 pass, and use 1246 Min, 2617 Max, and XXX Avg as per FitCD (I usually use 128 kbit audio files -- more on audio in a second). Next, I check the checkbox next to Upper Field First and set my Image Quality Priority slider to between 20 and 25 depending on the source material. I have often read that people like to use a number between 10 and 15 on the slider... Although this cuts down on the fuzzies during action scenes, it also tends to add them during the still scenes. Test it out and see what you like! :P

NOTE: If you are capturing a video converted from film -- most Hollywood movies for example -- you can include the IVTC plugin in your script. This will convert the output from 29.976 fps to 23.976 fps and give you 20% less frames to encode (less frames = better image quality). However, if you do this, don't forget to grab a copy of Pulldown.exe and apply this to the MPV file prior to multiplexing (next step). I haven't captured any film based video yet, so I am not sure how well IVTC works in Avisynth.

Now onto the audio portion... Once I have created my MPV file, I load the AVS script into TMPGEnc and encode the audio at 128 kbit using the tooLame external tool. (Yup, one step, that's it) Once that has completed and I have an MP2 file to match my MPV file, I multiplex them together using bbMPEG. Viola! A very nice looking MPEG2 file from your DC10+ capture.

Now, just as a comparison, here is my XVCD script I serve to TMPGEnc:

LoadPlugin("D:\DVD2SVCD\AVISynth\MPEG2DEC.dll")
SegmentedAVISource("D:\Test\Video1.avi")
AssumeFrameBased.SeparateFields
TemporalSmoother(2,1)
BicubicResize(512,224,0,0.6,16,8,608,224).Weave
BlendFields
AddBorders(16,16,16,16)
FlipVertical
ConvertToRGB
Trim(143,20632)

Pretty much the same script with a few small exceptions...
* I am using a resolution of 544 x 480.
* I use the BlendFields statement to deinterlace the video.
* I use the ConvertToRGB statement when serving to TMPGEnc.

In TMPGEnc, I select MPEG1, 544 x 480, 2 Pass VBR (1246 - 2000+ - 2617). I also encode the audio within the project... So there is no need to multiplex afterwards. The results are absolutely stunning. I was always anti-(X)VCD/pro-SVCD until I started using this resolution. And since the deinterlaced video compresses better, the above usually produces files about the same size as my 1/2 DVD SVCD's listed above at the same bit rate.

OK, I think that is it... I hope this has at least given you some ideas on how to use Avisynth with your DC10+ caps.

Good luck!

[Stripe]

Thanks Bewley

You helped me fix my wavey SVCDs , I was tearing my hair out on this one. God knows how many different test files I made using upper field lower- bottom- deinterlaced etc.

It was only when I read your post about the DC10 and cropping the image that I was able to fix the wave mpeg 2 picture using Premier avisynth and TMPGEnc.


Beware ALL Pinnacle DC10 plus capture card users . Capture at full res - do not check the TV box to crop image, this can be done by TMPGEnc

Cheers Bewley

[bewley]

Stripe,

I am glad that it helped.

Another thing for DC10 owners... It doesn't hurt to bump up the Sharpness slider a few notches when capturing. The DC10 card captures are really "soft" and this helps to make them a little more crisp.

[Bach]

bewley,
You are a true star. I have as many others done too many tests without finding that last little tweak. I'm sure this will get me on the right track again.
:P
PS. I do know how to use CCE, but I'll have to read up on Avisynth, but I don't really see that as a problem - finding the right settings is a lot harder.

[bewley]

Bach,

Setting up Avisynth should be a breeze -- if you have CCE 2.5 that is. If you are using CCE 2.62 or 2.64, then things get a little trickey (you will need to use VFAPI with ReadAVS.dll). I am more than happy to (try) and help if you have any problems setting up Avisynth or getting things to work.

Good Luck!

[Pitou]

Hi everyone,

I have a similar problem. I'll try to explain the best as I can.

In movement, the picture is like a Jello effect. It seems that some pixels don't follow the rest of the movement. So it makes the clip very, I would say, Jello!


I did resize and wanted to encode it in mpeg 2 using Mainconcept standalone encoder inerlace. I then burn it on DVD

Can this the cause of the problem??

Thanks.

Pitou.