De-interlacing workflow

[Jason J. Gullickson]

When using something like DVFilm Maker to de-interlace footage, do you usually do this immediately after capture and before you bring the footage into the NLE, or is it something you do once editing is complete?

[Joshua Provost]

Jason,

Deinterlacing is one of my first steps. If you plan to do any color correction or effects, it's cleaner to work with full, deinterlaced frames. My workflow is as follows.

1. Capture in Premiere (to DV-AVI)
2. Basic rough edit to remove bad takes in Premiere (to DV-AVI)
3. DeARTIFACT with Magic Bullet (to uncompressed AVI)
4. DeINTERLACE with DVFilm Maker (to uncompressed QT)
5. Color-correct in AfterEffects (to uncompressed AVI)
6. Edit in Premiere (to uncompressed AVI)
7. Finalize (opticals, letterbox, broadcast spec) with MagicBullet (to uncompressed AVI)
8. Encode to MPEG-2 in TMPGEnc
9. Author to DVD in DVD-Lab

Josh

[Jason J. Gullickson]

You have Magic Bullet but you use DVFilm Maker for progressive? Are the results better from DVFilm Maker (I had heard that Magic Bullet was the be-all-end-all of deinterlacing).

So do you go 60i to 24p or 30p, when you de-interlace?

[Joshua Provost]

Jason, it seems I am actually in the same discussion on this topic on a number of forums simultaneously. You can check out some of my arguments on this thread at pana3ccduser.com. It is in reference to this article at pvxuser.com, which concludes with the statement that Magic Bullet kicks ass. Now, mind the discussion forum attached to that article, because someone posted their own comparison highlighting some of MB's deficiencies there. The article prompted me to propose a better test methodology to the writer:

> Hey, this message is in regard to the 60i deinterlacing shootout.
> There is one major thing missing in your review and in all other
> reviews I have seen. Frames are compared side by side, but there
> is no "control group" or sanity check on what the deinterlaced
> frames should look like. What I'm saying is that the original
> footage should be shot with two cameras, one in 24p and one in 60i.
> The 24p should be the gold standard that the deinterlacers should
> be compared with.
>
> I bring this up because many have come under the impression based
> on stills that Magic Bullet is a great deinterlacer. Those of us
> who have used it and its competitors extensively have a number of
> walls with MB, and I personally have neve been able to get usable
> footage out of MB. It is always plauged with strobiness, twitter,
> noisy details, slurred patterns, or a combination (or all) of the
> above, due to the fact that it exclusively interpolates frames,
> leaving nothing in the original frame untouched. It creates stills
> that are very sharp and impressive, but actually overly precise for
> moving objects, resulting in the repeated complaints of it being
> stroby.
>
> Anyway, if you ever follow-up on this test, I would love to see a
> control camera shooting pure 24p in the mix.

You can demo all of these deinterlacers, so I'd encourage you to run some of yoru footage through them and evaluate the results on your own. However, be careful not to evaluate based on stills alone. MB tends to produce stills so sharp that the footage looks stroby on playback. Maker is less impressive as stills, but more pleasing in motion. And there are major problems with MB's interpolation that result in strobiness, twitter, noisy details, and slurred patterns, most of which are illustrated visually on the dvxuser.com respsonse thread to the article.

BTW, I go from 60i to 24p most of the time. Sometimes I go from 60i to 30p, then slow down to 24p for a slight slow-mo. Other times 60i to 60p to 24p for a super slow-mo.

If you go from 60i DV-AVI to 24p DV-AVI you are going through a recompress cycle, and will degrade quality. Doesn't mean you can't do it, but you should know what's going on under the hood.

[Jason J. Gullickson]

Originally Posted by Joshua Provost

Doesn't mean you can't do it, but you should know what's going on under the hood.

I definately want to know more about that!

I'm using Final Cut and I'd love to know exactly what's happening to my footage when it comes in from the camera and goes back out as Quicktime so I can avoid re-compression, transcoding, etc. It seems like I ran across something that covered this once but of course I can't find it now.

It sounds like a good place to start would be to import some footage, try a few of these de-interlacers to take it from 60i to 24p and go over the results in a "playback" environment simular to the way the final product will be displayed.

Since our work is entirely cinematic in nature we'll be targeting projection, most likely digital but hopefully 35mm transfer as well. That being the case, is 24p where I should be starting or is there no technical advantage (as opposed to asthetic or artistc) to a particular progressive frame rate?

[Joshua Provost]

Jason, being on Mac you have the advantage of a few more options than I do on PC, specifically Nattress and Apple Compressor. Do some tests on a variety of types of footage. Remember that deinterlacing implies motion, either objects moving through the frame, or the whole frame moving as in a pan or tilt of the camera. Take a good look at the footage they produce in motion, and look at how they handle details and patterns.

There is an advantage to 24p. It has a familiar quality associated with narrative storytelling on film, and being progressive it has more perceived resolution on playback and can be converted to film out. If possible, try to get a real 24p camera like the DVX100A or XL2. Even the best deinterlacing is a major compromise.

[Jason J. Gullickson]

For now I'm living with 60i aquisition (I just bought a Canon Optura 60) but down the road we'll be either buying/renting cameras that will be capable of such things...

Thanks for the info, I'll start experimenting immediately :)

[Joshua Provost]

Jason, I would love to see some side-by-side comparisons of a properly tweaked DVFilm Maker, MB, Nattress, and Apple Compressor, when you get into it.

[Jason J. Gullickson]

This is Wisconsin, and Winter (the other season) is coming, so I'm sure that in the following months I'll have many cold dark nights to spend on such a thing...

Thanks again for the help everyone and when I get some results I'll post them.

[Dave Perry]

Originally Posted by Jason J. Gullickson

I definately want to know more about that!

I'm using Final Cut and I'd love to know exactly what's happening to my footage when it comes in from the camera and goes back out as Quicktime so I can avoid re-compression, transcoding, etc.

Hello Jason. I use an Optura Xi, shoot only in 16:9, and use FCP5.

Regarding what FCP does to your footage when capture then exported as a QT file, it does nothing. That's assuming you capture via firwire and export using "QuickTime Movie", not "Using QuickTime Conversion". There is no transcoding or recompression going on. The .mov DV file is the same codec as the .avi DV file, it's just that .avi and .mov are different containers for the DV file.

Regarding 24p conversion, I use Nattress film Effects and love it. I also use Magic Bullet from time to time for the vignette effect it has.

Another advantage we Mac users have is the robust color correction available to us right in FCP. I think that is just as important to achieving a film look as deinterlacing.

[Jason J. Gullickson]

Originally Posted by Dave Perry

Regarding what FCP does to your footage when capture then exported as a QT file, it does nothing.

What about footage that gets "rendered", color correction, etc? Doesn't that require a decompress/recompression cycle?

[Joshua Provost]

Jason, absolutely right. If you change anything, titles, transitions, color correction, all of the effected frames get recompressed.

[Jason J. Gullickson]

Originally Posted by Joshua Provost

Jason, absolutely right.

Damn!

(sometimes I hate being right)

Damn! Damn!