View Full Version : converting interlaced footage to 24p with magic bullet
06-20-2005, 06:04 AM
I am currently studying to become a digital matte artist and I am looking to buy a prosumer DV cam to get some moving footage on green/blue screen.
I have been told that the dvx100 is the camera to go for because of 24p, but the Orphanage claim that if footage is captured on a fx1 (PAL) it can be de-interlaced and converted acurately to 24p.
I am looking to buy now and want some help.
DVX100 or sony hdr-fx1 with magic bullet??
Please help meeeeee......
06-20-2005, 09:09 AM
PAL is converted to 24p by deinterlacing from 50i to 25p, then slowing to 23.98fps. You can do that with practically any software, with great quality at a fraction of the price that MB costs. However, shooting high def and then scaling down to SD rez will give you better chroma resolution. Leaving it as HD will leave it with as equally bad chroma as DV, and higher compression too.
What edit system are you going to use the camera with? Do you want DV rez or HDV rez?
06-20-2005, 10:02 AM
Uummmmm, I would recommend the DVX100A as it seems to be a proven green/blue screen solution choice. I hear mixed things about the FX1/Z1 when it comes to chroma keying, some say they get terrific results and the rest say that they can't get a good key to save their lives and that they'd tried everything. Green/blue screen stuff is hard anyway and I just don't think the whole Sony HDV thing is working for that at this point (maybe JVC's HD-100 will do a better job). It also depends on the software being used.
Deinterlacing is good but you have to take into account your audio matching up and it just seems like too much hassle when you can get a DVX or XL2 to do straight 24p/24pA.
I think the biggest deciding factor would be dependant on your software solution, what are you using Daniel?
06-20-2005, 02:04 PM
hear mixed things about the FX1/Z1 when it comes to chroma keying, some say they get terrific results and the rest say that they can't get a good key to save their lives and that they'd tried everything
These people probably can't get a good key to save their lives off of DV footage either. Unless you can prove the DVX gamma and increased noise vastly improve keying, then basic math says you'll get better colour sampling out of an FX1/Z1U.
06-20-2005, 02:17 PM
How is very heavily compressed 4:2:0 better than heavily compressed 4:1:1? I just don't see that! Especially when the 4:2:0 comes not from real 1440 pixels, but from 960 magically uprezzed....
06-20-2005, 04:09 PM
Hey that looks like ATARI's type of math.
06-20-2005, 06:16 PM
Thanks for the replys.
I am using FCP 5 and will be using Shake for keying and compositing. I have used After Effects a few times but I feel much more at home in Shake.
I played with the demo of magic bullet and it does a very good job of deinterlacing and adding preset colour mixes, but Im using a G5 imac and it takes a while to render.
I was going to go down the DVX100 route but when I was looking at the 16:9 lense, it seemed that the list of things I would need was going up and up. And from what I've seen both the DVX100 and FX1 do a great job if they are in the right hands and the opposite is also true.
Basicly, at the end of the day, I just need something that will get me footage at a good enough resolution for a showreel that will be duplicated on a standard DVD, and produce enough quality that post/effects houses will take the time to watch it.
Plus shooting some weddings along the way wouldnt be out of the question.
06-20-2005, 07:11 PM
How is very heavily compressed 4:2:0 better than heavily compressed 4:1:1?
Hey Graeme, why don't you have a look and tell me which of these pictures is better:
http://s94963366.onlinehome.us/HDRFX1/softvincam.bmp (2 MB)
One is a down-rez from HDV to SD, the other is the in-cam down-rez to DV. Go into the colour space, and have a look at the top image. Tell me if you think a native DV acquisition can come close to the colour resolution and detail found there. Alternatively, you can see what 1 generation of DV compression does, by comparing the top image to the bottom one.
What's missing from this picture is a native DV acquisition. So obviously the quality of a) is better than b). However, since you're an expert at looking at DV footage, see if you can tell me that a native DV acquisition will be better, by illustrating what the horrible MPEG-2 artifacts in the top image are.
06-21-2005, 07:40 AM
Spiff, you're missing the point again. HDV as HDV is very heavily compressed 4:2:0, but when you downsample it in software to SD rez, you're making any compression artifacts 4 times smaller, and you're increasing the chroma resolution to 4:4:4, but the problem is, that what you get is not HDV any more, but SD. If you try to key on HDV as HDV you have exactly the same problems as you'd get keying on any high compression chroma reduced format, and indeed, even more so because 4:2:0 is even worse for keying than 4:1:1. By downsampling to SD rez you get rid of those problems, but you also loose your resolution.
So yes, you can get better keying off the HDV camera, but only by reducing it to SD rez, at which point it isn't HDV any more. Keying on DV in it's native state and HDV in it's native state was what I was comparing.
06-21-2005, 09:04 AM
I know what you're comparing Graeme - and I'm not missing the point. I'm comparing apples to apples here: SD to SD, and I've thought through the potential work-flows pretty damn thoroughly.
The point is, with the DVX the absolute best key you're going to get will be an SD key. With the HDV footage, and HDV resolution, you will have the 4:2:0 and other general crumminess for keying to deal with and will have to work with that for HD keying. HDV is not an optimal keying solution in HD to be sure... but if you're going to compare with the DVX, compare at SD resolution using the best work-flow you can.
Suppose you want an HD key.
With DV, this isn't even an option. You can key your 4:1:1 and scale it to HD.
With HDV, you can:
- key HDV in its native 4:2:0.
- attempt to reconstruct more resolution to 4:2:2 (or 4:4:4) with up-rezzing and sharpening filters to key at HD resolution (best solution for HDV)
- you can down-convert in cam and key 4:1:1 DV (idiotic - more artifacts, loss of chroma) and up-convert to HD
- you can down-convert to 4:4:4 in software and key 4:4:4 uncompressed SD with negligble artifacting and up-convert to HD (best SD solution)
We're working with 25 Mbps here. HDV is definitely a smarter use of that bandwidth. Furthermore, if you want to blissfully ignore the "magic" of pixel-shift and harp on progressive shooting, then fine. Using Cineframe modes you have a 960x540 4:4:4 image being stored in a 1440x1080 4:2:0 grid. There's more than enough space to get all that 4:4:4 in the decimated colourspace - and if you have any "faith" in the compressors at all, you should realize you end up with a pretty darned good 960x540 stream there.
By the way, my Swedish FX buddy (www.master-zap.com) who works for Mental Ray has very often said that 4:2:0 is a superior keying space to 4:1:1. Probably because the colour space smearing is more square than 4:1:1 manages.
The bottom line is that HDV will produce better keying results than DV. HDV is probably very comparable to DVCPRO-50 for SD keying, if not better.
People need to stop saying "compression is my enemy". HDV is more information more intelligently stored in 25 Mbps than DV, not less.
06-21-2005, 12:42 PM
Spiff, now we're singing from the same song sheet! HDV does indeed give more options, and when sensible downrezzed give better keying, that's agreed.
As for 4:2:0, because HDV is interlaced, the 4:2:0 is basically 4:2:0 on each field interlaced, which effectively makes quite a mess. 4:2:0 progressive is fine, as then it's 2x2 blocks, but on interlaced it's 4x2 blocks interlaced, which is tricky to deal with indeed.....
HDV surely does use a more effecient codec, probably on the region of 2.5 times as effecient, depending on program content, but it tries to store 4 times as many pixels for that data rate, so if you compare a 720x480 section of an HDV image against a DV image it will look inferior, but if you take the 1440x1080 and downrez it, it should look superior.
Compression is a fact of life, but you've got to understand exactly how it works to use it to your best advantage.
10-24-2005, 07:31 PM
ok... i'm starting to get dizzy reading all of this info...
i'm posting a music video shot entirely on green screen with an HDV sony camera (FX-1 i believe)
i am editing with Final Cut Pro 5 in HDV Native. now... i'm srarting to experiment with ways to key out the green.
what is the best way to do this with simple steps laid out please?
here is what i've tried so far with mixed results...
my first attempts were using Ultimatte vers 1.6 which I can run from FCP as a plugin.
This plugin is extremely render intensive, it looks like it gives good keys when i start testing it... but after rendering it's clear that the edges are a bit jagged and sloppy.. it's cumbersome to tweak because of the render times..
i've played around a bit with final cut pro's plug ins and these work faster but aren't as precise as ultimatte although i am more capable of tweaking them and checking the results.
so... what does everyone recommend???
10-24-2005, 09:47 PM
I've done some tests with HDV green screen and final cut and actually got really good results just using the final cut pro chroma keyer, much better than with DV footage. It's hard to say how much success you'll have because pulling a good matte depends as much on how the footage was shot as the program doing the key. If the DP lit the screen evenly, avoided fine clothing and hair and shot with a higher shutter speed to cut down on blur you should be in good shape, but there really is no standard way to pull a key. Every shot usually requires slightly different settings so it's always an experimental process.
10-25-2005, 02:39 AM
what is the workflow to downres a 16/9 footage shot with a fx1 in fcp5 please?
Sony FX1 User
10-25-2005, 05:41 PM
using the Sony fx1 for chroma keying with a green screen and it is very good.
i use the seem green screens as for the harry potter films and Alexander.
It is very in portend that you use good screens end lights.
look for green and blue screens on this website www.bristolpaint.com
10-25-2005, 10:28 PM
im thinkin of purchasing a fx1 and I am curious about deinterlacing to 24p as well. What I want to know is if anyone here has some sample footage from the fx1 that has been converted to 24p? Also, if I had a chance to purchase the pal version, would I be better off for conversions to 24p?
10-26-2005, 07:01 AM
No, no, no, Kyle, don't go with PAL. At least not for an FX1, an XL1S or DVX maybe, but not an FX1. Working with PAL in the US, an NTSC country, is a lot more strenuous than just using the NTSC version. By converting over to NTSC, you'd have to tweak colors, audio speed, and convert over to NTSC anyway, thus losing that extra resolution information via the conversion. If I'm not mistaken, those are the big problems with going from PAL to NTSC. I would suggest, if you're really adament about it, to go with a Z1U which is NTSC and PAL switchable. I've heard that Z1U's Cineframe 25 looks more promising than Cineframe 24, but I have yet to hear about the results and final verdict.
10-26-2005, 07:31 AM
Yes, Go Z1 which is PAL / NTSC switchable. Using PAL is fine, IF YOU KNOW what you're doing. Conversion / audio / etc. are not too bad, but you've got to understand the process. There should be no difference in colour as they're bot REC709. Yes, CF25 is better than CF24, but 50i de-interlaced in post looks best, and gives you the most options, like to do super-nice slowmo.