Finally! FX1 filmlook on par with true 24p

[Mark Kubat]

Hi folks.

Know this post/thread will eventually be moved to the editing section but I wanted to share the good news with you all so I thought I'd post here where a lot of you would see it:

we've done extensive testing this past week at our t.v. station with Edius 3.1 working with a rented FX1 (NTSC)...

The big news so far is we've found a really good way to get a true 24p type look as demonstrated by the DVX100/A... our method below beats CF24/CF30 hands down.

We captured native 1080/60i m2t's into Edius and edited our project. To make a final render, you have to use procoder express 2.0 for edius 3 which has a number of export options for HDV...

We selected to render out as 720/25p PAL.

Next, we imported this rendered m2t into Sony Vegas 5.0b to use Mainconcept mpeg2 (we could have stayed in procoder/Canopus but a lot of people here swear by vegas mpeg2 capability and it's been adopted by many here - we're more familiar with it, using custom 2-pass templates etc.) - we rendered out as 29.97p (progressive) mpeg 2, 2-pass, high bit-rate avg. 7.5...

Wow, the end result is truly amazing - very film like.

Okay, okay - I know people are going to ask - why 29.97p? why not 2:3 pulldown? what about audio?

This was just our first test going this route with edius out to 25p to see if it made a difference - sure enough, it does. The resulting 29.97p mpeg2 we created looks very clean - I don't know how canopus procoder takes 1080/60i and converts it to 720/25p PAL (ie. what algorithm, etc.) but the bottom line is yes, you can really really get nice 24p type effect from your 1080/60i footage going this route...

interlacing artifacts we saw in original mpeg2 derived directly from 1080/60i like "dancing" power lines and a lot of the motion artifacts typically acredited to HDV2 now seem to be gone - it's a very clean, progressive image - really, imagine that the FX1 has 24p type image and that's what our end result looks like! No cf24-type drop frame stutter crap! No cheap frame blending resulting in motion artifacts - this looks really good! In going to final mpeg 2 SD DVD, the detail is much higher/sharper and much less noise in low-light, higher gain settings... it is UNREAL how well the FX1 produces a CLEAN image in low-light...

We'll continue to experiment but right now just want to say this bodes well for us who want to do projects with the FX1 that look "filmic" - it is a better result than originally tried by those trying to use Vegas 2:3 pulldown with cf24 to get useable result...

The switchability of the Z1 is definitely going to be helpful... Hooray, hooray - the FX1 can definitely be used for indie filmmaking type projects!

[Dave Perry]

I was under the impression that the FX1 shot 24p any way, but admittedly, I know very little about it.

[Hayden Rivers]

Dave, the FX1 doesn't shoot native 24p.

Mark, any chance we're going to see some clips of this magic?

[Dylan Pank]

<<<-- Originally posted by Dave Perry : I was under the impression that the FX1 shot 24p any way, but admittedly, I know very little about it. -->>>

Dave, the FX1 mode Cineframe 24 sort of simulates 24 fps. It does this by discarding a field (to affect the "p"). How it gets the "24" is unclear but it seems that there is a certain uneveness about the frame rate in CF24 which implies it scans at 60hz, then either deinterlaces then drops one frame in 6, or it's choosy about which fields it's discarding when it de-interlaces. It then inserts a 2:3 pulldown to produce 29.97 fps, so every 3rd and 4th frame out of 5 will contain interlace artefacts.

It's a long way from what most people here would consider real 24p, i.e. what the XL2 or the DVX100 do, which is scan the whole CCD at 24hz, and then interlace that into 29fps, or what DVDs have which is real 24 frames per second with 2:3 flags.

[Mark Kubat]

What would be the best option for you guys to make you true believers? Okay, how's this? Is Chris Hurd able to host stuff here? I'll provide some m2t's if that'll do the trick - just want to confirm too that audio doesn't seem to be an issue. Guess it's adjusting automatically...

For me as an indie filmmaker, this right now is the way I'd go if I wanted to use the FX1 for a project. The indie t.v. station I work at broadcasts in SD, but we broadcast mpeg2 program streams (as is the norm for digital television/satellite) so I have to laugh that even m2t's are recognized natively in our software/apps here for broadcasting...

Re: clips - I guess some "before" and "after" stuff might be the order of the day?

Let's get some input from our wranglers here - maybe they can suggest something.

[Rob Lohman]

Actually, you would've been better of using the MPEG-2 encoder
in ProCoder. It is vastly superior to the MainConcept one in Vegas
and is currently the best software encoder out on the PC for an
affordable price!

The current list goes like:

1. Canopus ProCoder
2. CCE / TMPGEnc
3. MainConcept (Vegas, Premiere)

I also don't see how you got 24p since you got 25p in between
and 29.97p (or 30p) at the end. No 24p anywhere.

You are also doing double MPEG-2 compression (or actually triple
of you count the first pass in the camera). Why not output output
from Canopus ProCoder to 24p directly? Since the output to 25p
(according to your story) makes this look so good I would assume
it would make 24p look just as good since it is just a 1 frame
difference.

In the end it looks like ProCoder is doing a very good job on such
framerate conversions, which isn't too surprising to myself since it
has such a kick ass MPEG-2 encoder onboard.

[Hayden Rivers]

I think Chris has often offered to host clips, so that might be a good route. Some before and after stuff might be good, but that's 2x the bandwidth and I think I might rather want 2x the After.

So ask Chris. I've told Chris I think DVinfo.net needs a bittorrent tracker specifically for trading these clips. Hopefully one day it'll happen.

[Alex Raskin]

My experiments with quality deinterlacing in post concluded as follows: FX1's built-in deinterlacer in CF30 mode is better.

[Mark Kubat]

Well, it's good to see some discussion of this - clearly, there is an interest in trying to maximize the film-look potential of the FX1/Z1 - yes, I know a bunch of people will now post saying "but I love 60i" - well, I love it too but the producers I try to sell my projects too are trying to sell the final product to a market that goes for that film look... 'nuff said on the "philosophy."

Rob: I was suggesting that the end 25p matched the cadence/look of 24p DVX100 projects I've put out to DVD fairly closely (ie. identical) in terms of cadence, filmic look that is not the traditional 60i look. I didn't obviously get exactly 24p with what I've done, although I'm glad for the other responses suggesting how that can be done...

At our t.v. station we get a lot of PAL newsfeeds shot in 25 fps and to me it's just as good as 24 - I sometimes find the DVX 24p a bit too "stuttery" - more like an old 8mm home movie camera rather than the look of today's 35mm - 25p is good enough for us...

I guess the bottom line I'm trying to hit home is you can take 1080/60i that looks so "videolike" and do something nice to it via Canopus edius/procoder that gives you a filmic looking end result that in the end works waaaay better than cf24 ever could. You keep the full resolution of your source material and lose a lot of the artifacting due to m2t limitations I guess also tied in to the fact that it's interlaced - there is an appreciable improvement when you make a final SD DVD and I am 100% willing to say without a doubt that there is a superior end product compared to what the DVX100A gives you in 24p... sorry Barry Green! I do comparisons on a daily basis here at the station and everyone surveyed (including the UPS guy) picks the FX1 DVD over the DVX100 when asked simply "what looks better to you?" They go "what do you mean by better" and I say "well, just watch and tell me which one you prefer" - they make a call and pick the FX1 DVD EVERY SINGLE TIME... I ask why. They say stuff like "Oh, it's sharper. You can see more." stuff like that...

cheers!

Of course, when I showed it to them on an 8" CRT, they said there was no difference - that is an in-joke, folks!

[Chris Hurd]

<< Is Chris Hurd able to host stuff here? >>

Yes I am. Contact me offline via email.

[John Gaspain]

Its interesting news...im curious.

[Jon Fordham]

Mark,

I am not involved in post production and therefore don't quite follow your methodology or how exactly that methodology is producing a non interlaced motion signature with natural movement. More importantly, I am very curious as to how the workflow you've outlined "keeps the full resolution of your source material and loses a lot of the artifacting". How does the interpolation of the interlaced fields maintain the full vertical resolution? How does the resolution and clarity not take a hit when going through those multiple levels of compression and rendering? And how exactly do those multiple levels of compression and rendering remove the artifacting that m2t produces?

[Mark Kubat]

er, well, Jon....

did I mention that when we BLASTED our set with 1K fresnels we found that "latitude" of the FX1, like, sucked? But then my trusty camera assistant, "Beaker," pointed out that we were blowing out the image, afterall...

It was only when a couple of breakers tripped that we had 1/4 of the light and, voila, shots that looked better than "Collateral."

Cheers!

[Hayden Rivers]

I couldn't care less about another DVX, XL2, FX1 debate. I've seen plenty of DVX footage and it's great. I've seen some XL2 footage and it's pretty great too.

Now I want to see how far people can push the FX1/Z1. I'm sure 24p HDV or some 24p HD prosumer variant is going to arrive, but until then, I want to see the FX1 footage in question.

[Mark Kubat]

Well, I'll add to the report and say that use of Magic Bullet looks to the HDV footage in Vegas in going to mpeg-2 for DVD output looks very very nice - the big plus here is the very "clean" image you get from the Sony under low light conditions. A lot of the Magic Bullet looks work to increase saturation, contrast etc. and if you have low-light pixel noise in your mini-dv footage, using Magic Bullet actually will hurt you rather than help in many cases.

This is not so with HDV - it's really hard to get noise so in fact look presets like Color Reversal do a great job of treating your footage to give it the look of something that was acquired on film - in combination with the 25 fps methodology here, it works great... what can I say? Black is BLACK and you don't see pixels in low light... it's really nifty.

I'm preparing some materials to send to Chris Hurd so we can host them here. Give me a few more days (busy busy at work).

Cheers!