Dennis Couzin
Forum Replies Created
-
Heh! I sent a 1920×1080 8-bit uncompressed 4:2:2 file into Compressor for mpeg-2 compressor and the QuickTime player says the mpeg-2 is 720×404. The mpeg-2 looks great when played with Streamclip.
If you think about it, DCT compressed video has no inherent pixel count. Look at Figure 1 in Tudor.The 64 pixels in the image aren’t really represented by a 64 coefficient DCT block. Tudor’s text oversimplifies, while his figure is correct. The path shown snaking through the coefficients actually makes an increasing triangle based on relevance (roughly measured by u+v). The snaking path gets truncated in the compression, but a triangle is never a block.
The image pixels disappear in a DCT compression and other, non-corresponding, things take their place. In decompression, pixels are created. Why then is QuickTime player reporting 720×404? Does its mpeg-2 decompression algorithm always create 720×404 pixels, and leave further pixel creation to a scaling algorithm? Is that why it has even number 404 vertically rather than 405 (as 16:9 proportion requires)? Working with Apple software is a guessing game.
-
Alexander, thanks for chiming in, but I don’t see how the Nattress article you cite supports the point. The question is this: if we have a project in FCP and export a straight QuickTime file which we then drop into Compressor to do this-or-that, would we get better results going straight from FCP to Compressor to do this-or-that? I should have linked the Warmouth article that I quoted from: warmouth
-
What I described is merely 2:2:2:2:2:2:2:2:2:2:2:3 pulldown as described in the Wikipedia article on Telecine (with the 3 at the other end). That article is not hopeful on that sequence. Perhaps some randomization in the length of the sequence would help.
-
Gary, you need to debate your Compressor finding with Ken Stone who wrote in “Converting Frame Rates in Compressor” just four months ago: “Why use Compressor if FCP can do the job? Because, while FCP can do it with little hassle, the truth is, that Compressor can do a much better job of converting frame rates using Frame Controls which is based on Shake’s Optical Flow technology. Using Compressor does take extra work, so if you need a quick way to work with different frame rates in the timeline, and it doesn’t matter if the footage is a little choppy, use FCP.”
My background is in film. Some of my suggestions to Riana McKeith are based on the experimental results described under “fancy slow” in “Notes on Optical Printer Technique” (https://sites.google.com/site/cinetechinfo/). Frame rate conversion is a terrible problem for film precisely because there is no film method for image interpolation. Digital video can do this. It is obvious that the best way to convert 24 fps to 25 fps is to follow frame 1 by frame 1.96 (which is 96% of the way between frame 1 and frame 2), then frame 2.92, then frame 3.88, … then frame 24.04, then frame 25, then frame 25.96, etc. All these decimal denoted frames are interpolations. How the interpolating is to be done is non-trivial, and some methods will be superior to others. It’s also obvious that some styles of animation are easier to interpolate than live action is. Other styles of animation might not be.
You write: “…if it is truly animation, why is the ORIGINAL ANIMATION not being rendered out of the application that created it at 25fps instead of 24fps?” Come on, not every animation is created in an animation application with choice of render out framerates. An animation can be made from 24 separately made drawings (or other pictures) per second. Then, the problem Riana McKeith posted can arise.
-
Gary,
I think your advice:
“if this is animation, why even screw with this, have the render done at 25fps”
is almost the same as mine:
“For some styles of animation, frame interpolation can be nearly perfect. You can try Compressor’s rate conversion with its ‘best’ setting.”The question is whether FCP’s 24->25 render will look as good as Compressor’s or some other interpolation program’s. Is FCP’s method of 24->25 rendering documented somewhere? Is Compressor’s “optical flow technology” documented somewhere? I think the most effective 24->25 “render” for an animation will use original frame 1 and then create 24 new frames (1.96, 2.92, … 24.04) before using original frame 25, etc. That’s a lot of interpolation, so it had better be good.
-
Gary, thanks for the info on conforming.
Riana McKeith’s latest post makes it clear that converting, not conforming, is what is needed. -
Riana, thanks for the clarifying. Conversion without changing length was how I understood you in the first place, and what I meant in “of course you can convert from 24 fps to 25 fps. The question is whether you can stomach the result.”
What is the style of animation? For some styles of animation, frame interpolation can be nearly perfect. You can try Compressor’s rate conversion with its “best” setting. Who has a better interpolator?
If that produces a mess, your best hope is insertion of one frame per second into the original 24, but I previously overlooked a superior method. If your video is 24p then it can also be reckoned as 48p with 24 identical pairs of full fields. Now instead of inserting one new frame, which is a pair of full fields, into the 48 fields, it will be much less noticeable to insert one field at one point and another field approximately half second later. (The approximateness is useful to hide the intrusions.) Then you have created a 50p animation and you must keep it at 50p. You can make a trial of a few second of this by hand. FCP seems incapable of handling separate fields, but Avid can do it.
For example:
original frame 1 for 1/50 sec
original frame 1 for 1/50 sec
original frame 1 for 1/50 sec
original frame 2 for 1/50 sec
original frame 2 for 1/50 sec
original frame 3 for 1/50 sec
original frame 3 for 1/50 sec
original frame 4 for 1/50 sec
original frame 4 for 1/50 sec
original frame 5 for 1/50 sec
original frame 5 for 1/50 sec
original frame 6 for 1/50 sec
original frame 6 for 1/50 sec
original frame 7 for 1/50 sec
original frame 7 for 1/50 sec
original frame 8 for 1/50 sec
original frame 8 for 1/50 sec
original frame 9 for 1/50 sec
original frame 9 for 1/50 sec
original frame 10 for 1/50 sec
original frame 10 for 1/50 sec
original frame 11 for 1/50 sec
original frame 11 for 1/50 sec
original frame 12 for 1/50 sec
original frame 12 for 1/50 sec
original frame 13 for 1/50 sec
original frame 13 for 1/50 sec
original frame 13 for 1/50 sec
original frame 14 for 1/50 sec
original frame 14 for 1/50 sec
original frame 15 for 1/50 sec
original frame 15 for 1/50 sec
etc. -
Tom,
Golly, expert #1 says it “will greatly improve the quality” and expert #2 says “advantages, if any, are so negligible in relation to the substantial increase in render time.” What is a non-expert to think?(It’s possible that FCP has taken on parts of Compressor’s “frame controls” since 2004, so there’s no longer the great improvement in quality mentioned by Warmouth, but then there’s no longer the big difference in render time mentioned by you.)
Frankly, what scares me most is the “if any” in your answer. If we really knew how FCP worked, if FCP weren’t such a black box (with bugs crawling out) and poorly documented, there could be no doubt whether there is quality improvement or there isn’t. I’m not implying that Warmouth knows how FCP works and you don’t. Not at all. Warmouth could think he knows and be mistaken. Was Warmouth’s statement a deduction or was it based on experience?
I don’t think we should generalize about what’s too much time for a render process. It varies from no time as when a client is literally breathing down your back, to oceans of time as when an artist seeks the last ounce of quality and tries renders taking 50 or 100 times the video duration.
-
Jeff Warmouth, writing in https://www.kenstone.net back in 2004, pointed out the advantage of exporting directly from the FCP timeline to Compressor: “transitions, effects, titles, and other ‘renderable’ footage will be sent directly to the Compressor without rendering into DV or other codec. This will greatly improve the quality of the footage.” Is this no longer true?
-
John, reading your post I thought I’d written it. We’re in perfect agreement.
In fact everyone might agree with us. There’s an ambiguity in the word “convert” which might be what prolonged this thread. See my comments just above.