Rather than attempt to continue my research into the subject, which has left me dazed and confused... I shall turn to this braintrust for my answers.

I shoot with a Z1U. Virtually all of my content is viewed on computer.

I have been shooting Cineframe 30, and utilizing AspectHD to transcode. Results are good...but of course, the desire is to always try to do better.

Mixed in with the HDV footage are text pages and 2D mechanical animations. Everything is cut together in PPro2, and output via its Media Encoder.

The final outputs are WMV for windows users. For Quicktime, I output an uncompressed MOV, then output via Quicktime Pro. I've never achieved satisfactory results for Quicktime directly out of PPro2, plus... the presets in QT7 are very convenient.

Can I be doing better by shooting 1080i, editing in interlaced mode, then using a 3rd party plug-in such as FieldsKit to deinterlace a master output file?

There is so much conflicting opinion out there, but I feel it is more important to see what Cineform users are doing to achieve best results.

Finally... given all of the above, what would be the best workflow for downconverting a completed presentation to SD/DVD viewing.

Thanks all....

Everyone will have their own workflow and opinion.

My $0.02. I have a Sony HC1. It's raw 1080i footage smart-deinterlaces very nicely to Cineform 720p60. Working with that - i.e., progressive HD avi footage - might simplify things when you come to integrate your text and 2D stuff?

I encode directly from 720p60 to DVD MPEG using Procoder.

Hi Graham....

when you say "smart-deinterlaces".... are you referring to Cineform's preset for ingesting Sony's Cineframe format?

Sorry, that comment wasn't clear was it?

No I was meaning a software based motion-sensitive deinterlacing plugin. Fieldskit is I would assume an example of this. (I have not used it myself.)

The "smart" or "motion-sensitive" bit means that it only deinterlaces areas of the frame that are in motion, thereby retaining as much detail as possible. A "dumb" deinterlacer blends or discards entire fields, and so has a bigger hit on resolution.

Personally, I use the TDeint plugin within avisynth. There is also a Smartdeinterlace plugin for Virtualdub that's nearly as good.

I must state for the record that I am the "dumb de-interlacer"

Okay... now I see what you're saying. You capture with a interlaced Cineform preset, and do your conversion to progressive with an outside software.

Does your Sony offer Cineframe? I ask, because I am curious about the quality comparison between it, and the process you describe.

I am fairly certain that Cineframe is not discerning motion or static portions of the frame as it works.... so by default, it would seem that using a "smart" de-interlacer is going to be best.

The HC1 has a "Cinema mode" digital effect, which is somehow not quite the same as true Cineframe.

In any case, I did some resolution tests and there is about a 35% reduction in vertical resolution when you use it. I think from other posts on DVi at the time, there's a similar drop with Cineframe.

The Virtualdub/Smartdeinterlacer combo is freeware, so I'd suggest you give that a try, side-by-side with Cineform-extracted Cineframe, and see which you like better. There won't be a big difference in resolution, and the Cineform solution is a faster and more direct workflow, so you may decide it's fine.

The "classic" smart deinterlacer plugin for Vdub is #21 on this list: http://www.thedeemon.com/VirtualDubFilters/ - there are some others there as well.

Graham... thanks for your input on this.

I am trying out the FieldsKit deinterlacer right now.... veerrryyyy slow, depending on the settings. A lot of parameters that deal with temporal and motion. I will know more soon, and report back.

I'm using it within PPro2, but the option is there for After Effects as well. No idea if the rendering times are the same, but I would imagine so.

Thanks for the heads up and the link and analysis....very valuable.

I hope more chime in here for both of us to learn what others are doing.

UPDATE: FieldsKit does a fantastic job. Depending on what combination of the many parameters you choose, the render times can be very lengthy. My test was of some fluttering leaves in a strong breeze. I utilized FieldKit's motion detection and its temporal function of analyzing adjacent frames. A lot of processing while rendering to a CFHD clip. There was no visible loss of sharpness...and other than the progressive frames, indistinguishable from the original clip.

Ahhh, my fave subject... deinterlacing.

I shoot with a Sony Z1 1080 50i, ingest to Medium quality Cineform avi's.

I've tried out DVFilmMaker, Magic Bullet, Boris Continuum Complete Deinterlace etc, and this evening I'll be trying out Fieldskit.

Almost finished editing an 85min feature film, using PPRO CS3 (PPRO2 was crashing all the time) and Aspect v4.

I've thousands of clips in an 85 min film, so I don't have the energy to go through each clip selecting the right motion sensitivity for each deinterlacer - so I've been testing them on their default settings (remembering of course that HDV is Upper Field First).

Exporting out a 5 second clip with no FX applied takes 13 secs.
Boris deinterlacer applied in PPRO CS3 takes 2mins 22 secs
you think that's bad...
Magic Bullet Deinterlacer in After Effects, the same 5 sec clip takes 34 mins!!

DVFilmMaker is much quicker - so I'll retest all those tonight, along with FieldsKit and will post the results.

I think the best final workflow will be to deinterlace the footage used (possibly by creating a TRIMMED project and running DVFilmMaker batch through it) and replacing the current 50i footage with the 25p footage. Might work!

Oh yeh, back to the subject, Cineframe is crap if you have any panning or fast moving action. Preferably shoot interlaced and use a smart deinterlacer, as you summize... come to think of it, HDV compression on a whole is crap if you have any panning or fast moving action.

But if you're shooting training-style stuff that's gonna be highly compressed and viewed on PCs, then you'll be keeping movement down to a minimum anyway.

I'm getting horrible digital artifacts/video noise from footage shot at 0db - the foreground lit well, but the background, deliberately not lit (cos it's supposed to be dark!), is a mess of nasty swimming pixels!

My advice - if you haven't already shot your next big project, sell the Z1 and buy a RED! That's what I'll be doing if the wife wins the lottery next week.

That's exactly why I choose the JVC-HD100 to have true progressive frames from the start. It's a bit more expensive then the Z1, but you get a lot in return.

Is it true true progressive? I heard somewhere that some JVCs still used interlaced CCDs and deinterlaced on the fly.

The JVC is 720 though, which might look a bit more pixelly when blown up compared to the Z1 at 1080 (and if your smart-deinterlace, you're not losing any resolution apart from high-motion areas).

We're still comparing rotten apples with rotten oranges though - HDV is worth avoiding IMHO if you have the cash to avoid it. I've learnt my less the hard, and expensive way!

I think the best final workflow will be to deinterlace the footage used (possibly by creating a TRIMMED project and running DVFilmMaker batch through it) and replacing the current 50i footage with the 25p footage. Might work!

HDV compression on a whole is crap if you have any panning or fast moving action.

But if you're shooting training-style stuff that's gonna be highly compressed and viewed on PCs, then you'll be keeping movement down to a minimum anyway.

I'm getting horrible digital artifacts/video noise from footage shot at 0db - the foreground lit well, but the background, deliberately not lit (cos it's supposed to be dark!), is a mess of nasty swimming pixels!


Douglas.... I'm glad to see you join in this discussion.

Another option would be running a smart DI on your completed comp. It would be an awfully long render, but perhaps less time in the long run.

The artifacting you describe is something I have never run into. I have shot some severely lit scenes, and the dark areas have been very acceptable. No black stretch.

A technique that I sometimes use, which is not thought of too highly in the HDV community is the use of a light diffusion filter. In my mind, a very subtle softening makes compression much easier. My shots can include a lot of water surface area in the frame, and I've utilized this approach with great success. Water of course, offers one of the toughest challenges.... lots of motion, high sharp contract, no predictability, and extreme dynamic range of brightness/darkness.

Most people who comment on the filtering technique say that I am defeating the purpose of HD by softening detail. Ironic when you consider these are the same people who spend hundreds to simulate the blur and softening film!

Even unfiltered, I haven't experienced the pixelization you seem to be describing.... very odd. If you have a chance to throw on a Pro-Mist, or other similar filter (maybe a contrast reducer as well), you may be surprised at the results... but very subjective as we all know.

Please report back your opinions on FieldsKit!

DI'ing the completed comp isn't such a good idea - especially as I've letterboxed to 2.35:1 and re-framed (including some post push-ins using Scale, slow-mo etc), colour-corrected using Magic Bullet Colorista and film-looks from Magic Bullet Editors - deinterlacing on top of these changes wouldn't yield good results - it should be the first "effect" applied.

Those factors and the fact that motion sensitive deinterlacing gets really messy when you have a cut from one clip to another, just confuses the motion sensing - hence why Magic Bullet Suite recommend a Roll A and Roll B approach.

The idea of locking video edit, then creating a new TRIMMED project I reckon would solve this. I've tested the smart deinterlacing of TMPGEnc Xpress (Deinterlace Always, Interpolate - Animation 2) works well, and best of all you can batch process a whole directory of clips, saving the resulting progressive clip as the same file name in a different directory.

Then you'd just swap the names of the clip directories and re-open your project - which would load the prog clips, this I think should work?! You could even create a new PPRO proj 1080 25p, and import the old sequence, I not sure how fussy the cineform preset projects are.

I haven't tested yet, but the last thing I want to have to do is right click, Replace with Clip for everything on the timeline - or in the Project window right click, Make Offline, right click, Link Media to point to the newly deinterlaced clip!

Side-by-side I tested Boris Continuum Complete 4 Deinterlace, RevisionFX Fieldskit (both added as an effect) and TMPGEnc Xpress (imported after having processed) on an identical clip.

My test is a 2 sec clip where some blokes walk across camera with a nice brick wall and angry man in the background.

In motion these all pretty much looked the same. However, viewing a detail of a still lets us see just how well the DI'er coped with it.

Frame 1 is the untouched interlaced original, complete with combing!
Frame 2 is Boris Continuum Complete 4 Deinterlacer.
Frame 3 is RevisionFX Fieldskit 2.
Frame 4 is TMPGEnc Xpress.

Fieldskit has introduced some nasty digital artifacts which Boris hasn't, but I think TMPGEnc Xpress looks slightly nicer of the 3.

I'm using default settings for all the examples, except:
Boris Deinterlacer (setting Upper Field First)
FieldsKit (detect motion, 2 frames, point difference?)
TMPGEnc Xpress (Deinterlace Always - Interpolate - Animation 2)

By the way for rendering the 2 second clip, FieldsKit took 94 seconds, Boris took 55 seconds and TMPGEnc took 9 seconds. Have to stress though that I haven't exhaustively checked the resolution loss if any.

Methinks using an external DI'er is always gonna be quicker than using a Premiere or AE plug-in.

DVFilmMaker is also worth looking at, but the GUI is pants and needs some work - it's not obvious if you're by default adding Red and motion blur to your footage etc...

Artifacts are another discussion altogether, I don't want to hijack this thread - I might post some examples and some testing of the Boris Continuum Complete 4 Denoise plugin, it might save a couple of shots I'm having problems with.

Douglas... great job with the stills.

I see what you're saying about the processing of individual clips... that's how I should have described my thinking as well.... smart DI wouldn't work very well globally.

I didn't even realize that the TMPEGenc had a smart de-interlacer function... very good news, and time to give it a whirl. At least for that expense you would get two very important functions.

Is it capable of outputting a CFHD codec?

Very valuable research you have contributed... and it would be interesting to start a discussion about the artifacting you are dealing with.... perhaps my eye is not as discerning, or there is something at variance in our respective processes, because I have not had to deal with that issue on a serious level.

Thanks again for your contributions here.... it is definitely influencing my decisions.

Yes, TMPGEnc Xpress exports to Cineform HD avi - so this process should be virtually lossless.

In Output Format Selection, chose AVI file output, then pick the Cineform HD codec from the list - works great, and picks up the correct settings by default.

I've tried Sorensen and exporting to DVD etc in Premiere... and found TMPGEnc Xpress to make the best looking SD DVDs - so that's 2 functions I'll be using TMPGEnc for, great VFM.

Doesn't look like it handles HD-DVD or Blu-Ray encodes currently, but I'm sure it's on its way.

Thanks very much Douglas... you have really helped me out. I feel as though I have solved several nagging issues in one fell swoop.

I am going to get me a Foster's in your honor.

No-one drinks Fosters down here! Grab a VB instead.

Just tried my workflow on a tiny project.

PPRO CS3 project, Cineform 1440x1080 50i settings, imported 3 small CFHD interlaced avi's (reside in ProjectClips folder), added to timeline, saved project, exit.

Open up TMPEnc Xpress, Add clip, add Filter, Deinterlace (Deinterlace Always, Interpolate - animation2), chose Cineform codec, browse to new output folder, keep the same filename. Then add to encode queue - do this 3 times, once for each clip.

Encode. Then rename the original folder with interlaced clips to ProjectClips_Old and rename the new folder with the progressive clips to ProjectClips.

Re-open PPRO, et voila, the new progressive clips are used automatically - all effects controls etc should be still fine, as the clip length etc is unchanged.

If you're using Cineform, always make sure your Playback Settings are set to NOT De-interlace video shile scrubbing! Otherwise, you won't see the benefit, other than when you export.

I'm now happy with this workflow and will use it for the feature (so, I'll be removing the Boris Deinterlace fx from each clip and leaving the DI process as a post-edit process). I'll have to save a trimmed project before doing this though, as I have almost 1TB of footage - most obviously not used in the 85 min film... so I'll just be DI'ing the clips used.

-PS: Conforming Audio runs again for the new clips, so PPRO is obviously smart enough to know something's changed.

Has anyone compare the new MB 2007 with Fieldskit ? I saw a review of Fieldskit in this thread - but it is possible it used the default settings. The motion compensation settings need to be used for Fieldskit for any meaningful comparison;..
I think Fieldskit was marginally better than the old MB..
Just wondering if anyone has an opinion about the new version..
Thanks!

I'll not have time to more tests until I finish grading and start testing the rendering workflow.

But the tests I've done recently have given some very good results de-interlacing using TMPGEnc when I transcode my Cineform HD 1080i avi into PAL SD DVD.

I've bought the new MB Frames, so will probably use this anyway - but I'll test that soon and report back.

I'm sure both MB Frames and Fieldskit have trial versions, so if you're desp for a comparison, then there's nothing stopping you doing it.

Problems with a comparison is that there are usually clips in which one outshines the other. I am not exactly sure what to look for - so I was searching for the opinions of people who are "experts"..
At least more than I am..
I guess I will end up buying both MB and Twixtor.
Twixtor gives great slowmo effects and MB has the deartifactor, 24P and color grading elements..