Edit interlaced SD MiniDV capped files on HD progressive project....

[low-fat-al]

Any help with this would be great.

I have various old MiniDV and 8mm captured movies all in AVI (DV Codec). All interlaced of course.

I want a good work flow to edit on an HD Progressive sequence in Premiere Pro CS6.


Questions:

1) Prep footage first? Convert from interlaced DV to progressive? Best method?

2) Best method for upres or upsizing?


Any experts with great info is greatly appreciated.

[aedipuss]

best workflow is to keep it 720x480(or 576 if pal) and interlaced. maybe create dvds.

[sanlyn]

You're going to be disappointed. MiniDv never looks like HD. It can't. It just looks like upsized MIniDV.

Never resize interlaced video.

[Cornucopia]

I would use a combination of AVISynth script to:
1. Up-rez (resize up to HD)
2. De-interlace
3. Convert colorspace from BT601 to BT709
(not all necessarily in that order, probably in reverse order)
and then pipe that to an encoder that saves as Lossless intermediate files (in either QT/MOV or AVI or MXF container, depending on your and PProCS6's editing preference).

I'll leave it to the AVISynth script wizards to suggest the best options/settings, and these might differ depending on the exact settings you expect for your PPro HD sequence.
My guess is the resize would be something like Lanczos4Resize(), though a LOT depends here on how "sharpened" you want things to look.
My guess is the deint would be something like QTGMC(), though a LOT depends here on what kind of motion is in the footage and what your preference regarding blending.
My guess is for color space change you would use ColorMatrix(), but there are others (I don't do a lot of this type conversion, so don't have an expert opinion on it).

Scott

[Cornucopia]

@Sanlyn, I'm sure the OP realizes that, but it sounds like they want to combine older SD footage with HD footage into an HD edit sequence.
In that case, the only options are to downsize the HD to SD and do an SD edit, or upsize the SD to HD and do an HD edit. I agree than an HD edit is a better choice, future-proofing-wise.
Then it's a matter of: doing it natively from within PPro or using AVISynth/Vdub to prep. And in that case, if the OP is wanting as-good-a-quality-as-possible given the situation, I would say HANDS DOWN the AVISynth method is preferrable. And I believe you would say that also.

Scott

<edit> BTW, "never resize interlaced video" is a bit much. There are many good, pro ways of bobdeint->resize->re-interlace or bobdeint to 2x fps ->resize, or other acceptable solutions.

[sanlyn]

Probably better to downscale HD a bit and upscale SD a bit, and meet in the middle. What a headache.

[low-fat-al]

Thanks for the replies.
The footage is old skateboarding footage. So lots of action shots.
I want to put out a video for all my old friends from the footage but add newer HD footage (could compromise to 720p.)

I figured on deinterlacing and resizing using VirtualDub and AVISynth scripts. I want to export them as CineForm AVI's for editing.

Just not clear on the best deinterlacing scripts for this.... I messed around a bit but with no real luck in it looking great.

I can upload a short 10-20 second clip for someone to mess with perhaps.... let me know.

[low-fat-al]

I could even stick to Standard Definition if needed. ... just a passion project is all.

[poisondeathray]

Are you in NTSC land or PAL land ?

What is your HD footage? 1080p59.94 or something else ?

How much are you planning to use HD vs. SD source footage split ? eg. 90/10 ? 50/50 ?

How are you going to watch it ? blu-ray player ? Computer ?


My opinion: I would use 720p59.94 (or 720p50 in PAL land) , because I wouldnt want to degrade the HD footage too much, and it captures the motion smoothness if you bob deinterlace the SD footage . Best overall deinterlacer would be QTGMC

[low-fat-al]

Here are two files. One is from an old 8mm tape and the other from a newer MiniDV tape.

[_Al_]

Properties of HD video are important too, what are they? 29.97i, 29.97p, 30p or 59.97p , video size.

[low-fat-al]

I don't have any HD yet... was going to work on that after I went through the SD footage. All SD footage is NTSC

[_Al_]

I loaded clips into NTSC DV project in Vegas, used denoiser, exported with debugmode frame server, and loaded Avisynth script to x264encoder. You'd use Premiere and VirtualDub instead of Vegas and x264encoder to get some lossless (like ut codec or huffyuv) and you'd load those clips back to your main HD project.

Denoiser is a must because otherwise upscale would not be pretty especially with that 8mm source. I do not know how to use denoiser directly in Avisynth so I went through videoeditor, saves some hassles too and you can quickly set a profile for particular clip in editor with Neatvideo. You'd need only that family version, light version of Neatvideo or how they call it, because of SD size of your source. So scripts would be something like this:

720p video from NTSC DV:

avisource("D:\your_clip.avi")
#denoise here or after QTGMC, not sure
assumebff()
QTGMC()
nnedi3_rpow2(rfactor=2)
Spline64Resize(960,720) 
AddBorders(160,0,160,0)

1080i from NTSC DV video:

avisource("D:\your_clip.avi")
#denoise here or after QTGMC, not sure
assumebff()
QTGMC() #double framerate
nnedi3_rpow2(rfactor=4)
Spline64Resize(1440,1080) 
AddBorders(240,0,240,0)
AssumeTFF()
SeparateFields().SelectEvery(4,0,3).Weave() #makes interlace

http://files.videohelp.com/u/198160/1280x720 60p.mp4
http://files.videohelp.com/u/198160/1920x1080 30i.mp4

[low-fat-al]

Thanks for your reply. I can't view your samples. It says they are deleted.

I am going to try your script idea and see what I can do with it.

[_Al_]

try now:
http://files.videohelp.com/u/198160/1280x720p.mp4
http://files.videohelp.com/u/198160/1920x1080i.mp4

[low-fat-al]

Amazing results with your samples! I am curious what settings you have for Neat Video. I do use Neat Video as a plugin with VirtualDub.
Also I usually crop and resize in VirtualDub. Do you suggest I do the resizing and cropping instead in the script like you have?
Thanks again.

[_Al_]

In Vegas it is pretty straight forward , you select frame, then choose some homogeneous area and click auto profile. Not sure about VirtualDub plugin.
Upscale is here discussed from time to time - nnedi3 upscales original video always two times or for times and then following downscale to desired resolution supposedly brings better result than simple Lanczos upscale so I used nnedi3. Not sure how you do this in VirtualDub. As default, there is just simple Lanczos resize within resize filter.

But of course you can load Avisynth script into VirtualDub, to be clear.

[low-fat-al]

Ok. I am experimenting with a short clip with given example script with some modifications. (note: I removed the black side bars. It ends up as final resolution of 960x720. When editing, premiere could automatically fill the right and left areas or I could add an effect behind the video, etc...)
The scripts allows for proper clean deinterlacing, resizing, etc. I load it into VirtualDub and use the NeatVideo plugin. It works the same as the Vegas version. I also have the version for Premiere Pro. Anywho... I denoise it with that, etc...
I export using CineForm Codec set to High.

Final output is a video 960x720 @ 59.94fps.

So when I edit this with other footage, would it look alright on a sequence with other footage at different framerates? 59.94fps for the full smooth motion for these converted videos, but how does that translate to project. Final could be a web distributed video H.264 and highly unlikely a DVD or BluRay.

Curious on the project settings if source footages are mixed in frame rates?

Thoughts anyone?

[_Al_]

Personal choice of yours,
Shooting 30p (29.97) with HD camcorder and mixing it with upscaled 30p (29.97) footage (just add selecteven() after QTGMC() line) in Avisynth script for that older material.

I do not know about this, times go on, and you are stuck with tons of 30p footage for the future,..., it is your personal choice

I'd say shoot new material 60p (59.97) HD, mix it with upscaled 60p (59.97) material , set project as 60p, export 60p(59.97) as a back up, intermediate , whatever, and export 30p 29.97) for web. Not sure about Premiere, but Vegas will just drop every other frame (or use frame server to make sure) .
You can help it to look better shooting with longer shutter speeds (but not have it blurry) experiment with 1/60, 1/120 shutter speeds on your new camcorder ....

[low-fat-al]

I did a test on and SD set up. I took the old footage, exported one as 59.94 fps and the other with the "selecteven()" added in making it 29.976. I created a sequence in Premiere Pro with 29.76fps settings and another with the 59.94fps settings. I was surprised. I put both along with the original MiniDV captured clip in the timeline as separate video layers so I can have the ability to toggle on and off and see the same video frame. I advanced frame by frame and it appeared kinda jerky... like it was missing frames. This was in the 29.976fps sequence. I did the same thing in the 59.976 sequence and it actually had played the missing frames making for smoother motion. This was true with the original MiniDV footage and the converted footage at 59.94fps. The "selecteven()" footage at 29.976fps had every other frame.

I had no idea that I was actually missing frames even if I used the original MiniDV footage in a sequenced the matched it's settings as footage. Learned something new.

Now if I had to make this an SD project at 59.94fps (best for smoother motion) and I needed to make DVD's (NTSC 29.976i standard settings), would I lose the smooth motion in the conversion process?

[_Al_]

You had it right with original 29.97i mini DV temporal resolution is ok, 50 informations in one second about some movement

the whole point of using QTGMC is that it can calculate full frames out of those half frames that mini DV uses, so you have smooth and progressive footage before you upscale, only then upscale is quite alright, you cannot upscale interlaced video without getting some weird artefacts, so video gets progressive before upscale, but double frame rate so that temporal resolution is not lost

but skipping every other frame to get 29.97p could be troublesome, especially in videos where something moves - your videos, so it is best way just to stop right there after qtgmc and keep 59.97p

you can skip qtgmc() and use tdeint
TDeint(order=0)
to get 29.97p out of 29.97i where videos will sort of "blend" frames so it looks somehow smoother but on an account of not being crisp and smooth which is not good for upscale that follows, overall upscale is worse

Although, there is something like motion blur that you can set using QTGMC to drop every other frame so those gaps are not that obvious:
http://svn.0x00ff00ff.com/mirror/http/www.mediafire.com/su7l5jtcobabksk/QTGMC-3.32.zip...ml#ShutterBlur
that might help you simulate longer shutterspeed, you experiment with value bigger than 180, I tried it once, not extensively though , and did not like it, maybe you will get better results

but I would not do it anyway, as I said, just working with 60p, for the future sake