IVTC Virtualdub and TMPGenc

[grunteled]

Ok,

I have captured a LaserDisc to DV using the ADVC-100. Now I looked through the AVI that was created and I can see the telecine pattern. 3 progressive and two interlaced frames repeat over and over again. I have heard that it is best to convert back to 24fps by runing the IVTC filter to recover the orginal frame rate.

Fine, I use the Telecide beta filter for VirtualDub and some noise reduction filters after that and save the results in Huffyuv format. I expected to find an AVI running at 24 progressive frames per second. What I get is 30 progressive frames per second. That I was not expecting.

Loading the file into TMPGenc, it insists it is 30fps (29.97) INTERLACED. I can change that to non-interlaced but the results are not affected. The framerate of the final MPEG is 29.97 with no pull-down. The motion in panning scenes is awful and I can find no way to get the frame rate converted to where I want it... (23.976).

Am I missing something. I thought IVTC was supposed to leave you with 24fps progressive.

Thanks In Advance... G

[DJRumpy]

In virtualdub, you can perform IVTC by selecting VIDEO | FRAMERATE. In the bottom section "Inverse Telecine (3:2 Pulldown Removal)", set the option for "Reconstruct from fields - adaptive", and click OK. Save your new AVI, or frameserve it. The output will be 23.976 fps.

If you want to do the same in TMPGenc, select the 'Inverse Telecine' filter on the Advanced tab. You should also make sure you set the proper settings under the VIDEO tab:

Encode Mode: 3:2 pulldown when playback
Frame Rate: 23.976 (internally 29.97 fps)

Note that the Frame Rate setting may not be available until you set the encode mode.

Your output will be converted to 23.976, but will appear to be 29.97 to any authoring software.

[KetchSumAir]

@grunteled

That Telecide beta filter works. Are you doing that first, before the other filters? It looks like you are, and typically, you'd want IVTC first, IMO. The other option is to have TMPG look at it - perhaps the IVTC gets confused due to the 3:2 pattern changing mid-way into the film. TMPG has a auto mode in it's IVTC. If you use that, then click "check", it should show you if there is a change in the pattern that it can't handle. Check out this article:
http://www.doom9.org/index.html?/mpg/tmpg-ivtc.htm

If you start using options like "Reconstruct from fields - adaptive", you're actually de-interlacing. De-interlacing results in a loss of quality, versus strict IVTC which just removes duplicate fields.

For more info on De-interlacing, visit http://www.100fps.com

[DJRumpy]

If you start using options like "Reconstruct from fields - adaptive", you're actually de-interlacing. De-interlacing results in a loss of quality, versus strict IVTC which just removes duplicate fields.

Not quite accurate. 'Reconstruct From Fields' does not deinterlace your video. It uses the same IVTC process used by TMPGenc. Both look for fields that are duplicates, and removes them. If they cannot find a match (duplicate), then they will decimate the field count regardless. This is true for both TMPGenc, and VirtualDub. Deinterlacing is a completely different animal, which is not possible from the VIDEO | FRAMERATE menu in Virtualdub.

As to filter order, you should always perform IVTC first. Changes to each individual field can cause the IVTC process to fail, if the fields are changed enough that they no longer appear to be copies of each other. Noise filters and such are a very bad idea, as is resizing, or even cropping before performing IVTC.

[KetchSumAir]

@DJRumpy

I'm still a little new to the IVTC/De-interlacing concept. I thought I had it figured out, but your statement doesn't seem to line up with 100fps.

Could you set me straight?

Details:

The 1st reply you sent had:

Reconstruct from fields - adaptive

Your 2nd reply when you corrected me just called it:

Reconstruct From Fields

Now, I have admittedly never used that particular feature in VirtualDub you are talking about. However, with the word "adaptive", I assumed that meant what is reffered to at 100fps.com as:

"You could call it:
- Progressive scan
- Bob+Weave
- Intelligent
- Adaptive
- Hybrid"

If I understand the article, adaptive is a term describing a type de-interlacing, not IVTC.

Am I missing something? Did you mean "Reconstruct From Fields" or "Reconstruct From Fields - adaptive"?

I'm just a bit confused right now....

[DJRumpy]

I was referring to the 'Reconstruct From Fields - adaptive' option in VirtualDub. The 'Adaptive' description can apply to both deinterlacing filters, as well as IVTC filters. The word simply means that each filter type can adapt to changes in the video stream. For deinterlacing, the adaptive type deinterlacer will ignore content that is not or interlaced (e.g. progressive frames will have no combing affects). It will deinterlace frames that do have combing evident, adapting to each frame as it processes the video, hence the 'adaptive' description.

For IVCT, the word adaptive descrives an IVTC filter capable of detecting changes in the telecine pattern. Most videos will not be a strict 3:2:3:2:3:2 pattern throughout. They change when scene's change in the video, when edit's were done in post production, etc. With an adaptive IVTC filter, it can adapt to changes in the pattern. It does this by using the 3:2 pattern as a reference, but not a rule. If it looks at a field that is supposed to be a duplicate, and find's it does not match, it will then do a pattern scan, to see which fields (if any) match it. If none do, then it will usually do a full scan of 5 frames to see which fields are duplicates.

[KetchSumAir]

@DJRumpy
Thanks for the explanation - it make perfect sense now.

Final question: Have you tried the Telecide beta filter grunteled mentioned?

If so, I take it you like Virtual Dubs built-in one better?

I've admittedly only used IVTC on a few movie backups so far (from VHS), but have not had any problems with the Telecide beta filter.

Just curios as to your expierience.

Thanks.

[grunteled]

DJ and Ketch,

Thanks for your replys. I see this is an area of fairly great confusion for most. Good, I don't feel so alone. Anway, I took some of what I got from you guys and some of my own experiments from last night and have come to the following:

1.) The Telecide *FILTER* is not the way to go for NTSC. The author of the filter seems to aknowledge that is creates a wonderful progressive stream... just does not decimate frames. Start at 29.97 then you end at 29.97.

2.) The correct location does indeed seem to be under VIDEO | FRAME RATE. However, it does not like having noise reduction filters and such applied when runing IVTC. I seemed to get garbage everytime... at the proper framerate mind you, but garbage none the less. Once I ran the IVTC alone I got a good 24fps movie with very few artifacts.

3.) I HATE TMPGenc's IVTC. It produced a wicked mess each and every time and made me wait forever to get my mess. It just could not seem to hit the pattern (or I can't find the right settings) for nothing.

4.) Massive Disk Space Required. I have a 170GB capture drive and I'm feeling cramped now. 30GB for original capture. 90GB for Huffyuv version that has been IVTC'd. Another 34GB a piece for each segment to have filters applied to the IVTC'd film and stored again as Huffyuv. Sweet Jesus, I thought 200GB would be enough for final product and editing.

Why don't I just frame-serve? Well, when I do VBR encoding I have found that as TMPGenc runs through to look for motion, all the noise filters and such are applied (slooooowwwww). Then they are of course, applied again when the actual encoding happens. (damn slooooowwwww).

Other than being pinched for disk space I think I have a workable system here now to get the film encoded in 24fps. I'll let you know what I come up with when I get around to actually encoding the video. Thanks for all the input though. I might not have found the frame rate conversion without a little nudge. Thanks again!

G

[KetchSumAir]

@grunteled

Unless I'm miss-understanding something, you are using TMPG noise filters. I've found Virtual Dub filters to be MUCH faster, and better as well. This would mean re-reading the stream into Virtual Dub a 2nd time, but then you could frameserve to TMPG....

Anyway, just a thought.

[DJRumpy]

Final question: Have you tried the Telecide beta filter grunteled mentioned?

I use AVISynth to perform IVTC. I started off using VirtualDub to IVTC. AVISynth can perform IVTC, in addition to other filtering without having to do those other filters as seperate step/temporary file. If you want to continue to use VirtualDub, I would suggest you look at the Decomb fitler by DGraft ( http://shelob.mordor.net/dgraft ). It's one of the best I've found, and available for both AVISynth, and VirtualDub.

1.) The Telecide *FILTER* is not the way to go for NTSC. The author of the filter seems to aknowledge that is creates a wonderful progressive stream... just does not decimate frames. Start at 29.97 then you end at 29.97.

It sounds like it's just performing a deinterlace on the video. Any proper IVTC would reduce the framecount (decimate). I would find another filter (see the link for dgraft above).

2.) The correct location does indeed seem to be under VIDEO | FRAME RATE. However, it does not like having noise reduction filters and such applied when runing IVTC. I seemed to get garbage everytime... at the proper framerate mind you, but garbage none the less. Once I ran the IVTC alone I got a good 24fps movie with very few artifacts.

This is a known problem for VirtualDub. Changes to the individual fields to to fitlering give the adaptive filter fits. I would suggest you use a 3rd party filter like DECOMB, or my personal suggestion, use AVISynth + DECOMB

3.) I HATE TMPGenc's IVTC. It produced a wicked mess each and every time and made me wait forever to get my mess. It just could not seem to hit the pattern (or I can't find the right settings) for nothing.

Always do small pieces of clip using the Source Range filter under the Advanced Tab. It will save you hours/days of encoder induced stress. When your happy with your sample, then uncheck the source range. You should also look at the other encoders. They offer much more speed for comparable quality.

4.) Massive Disk Space Required. I have a 170GB capture drive and I'm feeling cramped now. 30GB for original capture. 90GB for Huffyuv version that has been IVTC'd. Another 34GB a piece for each segment to have filters applied to the IVTC'd film and stored again as Huffyuv. Sweet Jesus, I thought 200GB would be enough for final product and editing.

Why don't I just frame-serve? Well, when I do VBR encoding I have found that as TMPGenc runs through to look for motion, all the noise filters and such are applied (slooooowwwww). Then they are of course, applied again when the actual encoding happens. (damn slooooowwwww).

Frameserving shouldn't be slowing you down that much. It's definately the way to go regardless. Filtering is more likely to be the reason for your speed woes. TMPGenc is one of the slowest encoders on the market. Filtering just rubs salt in the wound . Check out Main Concept, or CCE. Both have cheap comparable products that will approach or exceed realtime encoding speed with filtering and frameserving. It will save you time, space, and age spots.

[grunteled]

Frameserving shouldn't be slowing you down that much. It's definately the way to go regardless. Filtering is more likely to be the reason for your speed woes. TMPGenc is one of the slowest encoders on the market. Filtering just rubs salt in the wound . Check out Main Concept, or CCE. Both have cheap comparable products that will approach or exceed realtime encoding speed with filtering and frameserving. It will save you time, space, and age spots.

Well the filters I'm running are in VirtualDub as it is frameserving. The noise reducers are all slow and usually bog down to 5-10fps even saving to intermediate AVI. I just hate paying that price twice when doing VBR from VirtualDub with filters on the VD side. TMPG's noise filter doesn't do so good for me with LaserDisc.

I'd buy CCE or Procoder, but my wife might start asking why I don't just buy the DVD... given the amount I have already spent. To which I have no good responce (other than stubborn headness, I refuse to pay for the movie a third time for a new format).

[DJRumpy]

I'd buy CCE or Procoder, but my wife might start asking why I don't just buy the DVD... given the amount I have already spent. To which I have no good responce (other than stubborn headness, I refuse to pay for the movie a third time for a new format).

LOL They always ask those 'sensible' questions don't they...

If you don't want to get bit by filtering on the second pass, then you have to save your filtered file as an AVI, and then feed that to TMPGenc. It's not worth the time with the other encoders, but it may be with TMPgenc.

[grunteled]

Thanks everyone for the suggestions and thoughts. Using the info from this site I was finally able to get my first (good) conversion done. Robin Hood came out very well. A little too bright, but not bad at all. It runs rings around my first attempt (Red October). Now that I have a system maybe I can start to pump them out more efficently.

Thanks again everyone!!

G

[DJRumpy]

You can use the 'Simple Color Correction' filter in TMPGenc to darken or lighten your encoded mpeg. The title for it is a little misleading, as it does more than just color. Make sure you do a small test clip first to ensure everything looks alright. Just something to keep in mind for your next project