Compressor 2 and 3 have very powerful tools for converting and upscaling video, as well as doing slo-mo. The optical flow technology, which Apple acquired when they bought the powerful compositing application Shake (only later to kill it), is the key to the results that you can get from Compressor. The key to getting the best results is the Frame Controls settings in Compressor.
Compressor is very good at standards conversion or converting video to a different frame rate. People often call this PAL and NTSC conversion although this isn't really accurate, but it will handle the same tasks that systems that do this do--namely the scaling and frame rate conversions. Frame rate conversions in particular are difficult to do really well. There are some standalone products from Teranex that just do this for around $2500 Well Compressor does this really well, but Teranex does it in real-time and Compressor is really slow.
Optical flow can also help in up-conversion of interlaced footage. Interlacing makes it really hard to do upconversion because your vertical resolution is basically split in two. For an example of how to scale things up badly look at what Final Cut Pro does in the timeline. FCP basically throws away half of your vertical resolution as soon as you scale a video clip. So the difference between 100% scale and 99% or 101% scale is crisp video and a blurry messy. When scaled up like this over 200%, not unusual when putting in standard-def footage into a high-def project, fine detail like type also starts shaking up and down from the interlacing--it looks ugly.
Compressor is an improvement on how FCP works in a number of ways. First, it can analyze the image and decide to selectively keep detail from both fields when the video isn't moving--if visuals in a part of a frame is moving, then only that part of the frame will lose resolution. With the optical flow technology, Compressor guesses how a part of the frame is moving and intelligently creates an image with more resolution than other results. These two approaches are handled in the deinterlacing settings in Compressor--the first approach is described as Better while Best uses optical flow.
The Retiming controls are where Compressor handles different frame rates. For example going from 25 fps video from Europe to 30fps video over in North America, usually handled as 60i (interlaced) especially in standard-def. The Better and Best settings in Compressor use optical flow. Like with deinterlacing, optical flow analyzes your video and recognizes object and how they are moving--the optical flow then intelligently creates new frames that move objects smoothly without losing detail. By the way, going from 24fps, which is a progressive format, is not a problem for us folks working in 60i. You can do this right in FCP without any fancy conversion (assuming you want to keep the 24fps filmic look.)
As an example, here's what I recommended to someone doing PAL to NTSC conversion:
Resizing: Better
Deinterlacing: Better (motion adaptive)
Retiming: Better
In this case, the scaling is going down (PAL has more resolution than NTSC, so you should get good results without jumping through a lot of hoops.) I reserved the optical flow use to only retiming at the Better setting. I didn't recommend optical flow on deinterlacing to save time. But if time isn't an issue then using Best for deinterlacing should improve results.
Unfortunately, all this intelligent analysis by optical flow is very processor and time-intensive. So for example using optical flow for retiming and and deinterlacing makes my quad-core G5 slow to a crawl--I'd give a ballpark estimate of 1 minute to process 1 second of video. But if quality is important, then the results can be excellent.
There are some plug-ins that specialize in upconversion, namely InstantHD. Check here for my post on upconversion.
Here are two good articles on using Compressor. Macworld discusses standards conversions here that walks you through the whole process:
Send video abroad
This post at Digital Rebellion gives detailed info on all the settings in Frame Controls.
Using Frame Controls in Compressor
The Macworld article also recommends an interesting workflow using Cinema Tools to convert PAL to NTSC by conforming the PAL to 24fps, scaling and deinterlacing in Compressor, then adding pulldown in FCP. Interesting workflow that I've never seen used.
The Retiming controls in Compressor can also give you some very good slo-mo results. Here are some cool tests done that compares Final Cut Pro versus Compressor at various speeds on two shots.
"Final Cut Studio slow motion demo"
"Final Cut slow motion at various speeds"
The author brings up a good point--if you are shooting a shot to be slo-mo, then shoot at as high a frame rate as you can--60p, 60i, 50p, 50i are good, where 24 fps is bad for slo-mo. The more samples per second of the image you can acquire, the better the slo-mo will work.
Showing posts with label standards conversion. Show all posts
Showing posts with label standards conversion. Show all posts
Friday, July 25, 2008
Sunday, May 11, 2008
24p vs. 30p
So we are moving away (slowly all be it) from interlaced video to progressive video. There are a number of framerates common with progressive. Here's a brief breakdown
24 fps--what's used on film
30 fps--half as many samples per second as regular TV (e.g., 60 interlaced)
25 fps--half the frame rate of regular TV in Europe (e.g., 50 fps)
50 fps--the samples of images per second used in European TV
60 fps--the samples of images per second used in North America TV ( which is usually interlaced)
So people generally prefer the slower frame rates in documentaries and dramatic material because it reminds people of film and the cinema--in short it looks classy. The faster frame rates look like TV and are a must have for sports where things are moving fast.
This post is primarily concerned with the technological differences (and not aesthetic) between different frame rates. (If you want to brush up on your basic knowledge of interlaced versus progressive, check out their definitions from this glossary from Quantel.) So let me outline a couple of the technical advantages of progressive video versus interlaced video:
1) Monitors are rapidly turning progressive--LCD's (both TV's and computers) and plasmas. These monitors can't natively display interlaced video. So some processing has to be done to de-interlace, and very often you end of losing resolution (and lots of it) when the de-interlacing happens. It depends partly on how good the hardware is that does the de-interlacing, but this happens at the consumer's end--so you never quite know how it will look. In short, a cheap TV will do a bad job of this.
2) Better effects work on progressive footage. Lots of effects, including scaling, require the video to be de-interlaced. Final Cut Pro does a lousy job of de-interlacing. If your footage is progressive (and FCP knows it), then it won't de-interlace to do simple things like scaling up the video. This past week, I had to include some letterboxed standard-def video in a high-def show. It looked super soft in FCP. So I used Compressor to do a very slow but very good de-interlacing of the footage, which I then scaled up (but again I didn't even scale this footage in FCP). It looked a lot better with the de-interlacing that Compressor does than the de-interlacing that FCP does.
3) Progressive distribution of video. Computer monitors were never interlaced--so anything for distribution on computers benefits from being progressive. Also DVD's will handle 24 fps. This is great if the video is going to a progressive display, e.g., LCD or Plasma. As mentioned above, interlaced footage needs to de-interlaced to be scaled, but progressive doesn't. So a 24fps DVD will usually look much better on a high-def display than a comparable 60i video.
Important note: DVD doesn't have a standard for 30 fps--it's either 60i or 24 fps. So your DVD player and your TV treat 30 fps just like 60i--if it gets scaled up then you could lose resolution (but not always--again it depends on the display.)
Now let me move on to outline some advantages and drawbacks of different framerates:
24 fps: So if you're going to make a film print, then you should shoot in this framerate. OK .001 per cent of anything shot by professionals will ever end up on film. But because 24fps has been very hot for many years (essentially the Panasonic DVX-100 set that off) there are a lot of post workflows for working with 24 fps. 30 fps is gaining ground though--there are now lots of setups in FCP for it.
24 fps is very close to 25 fps. So a general option for converting for a European master is to just speed up the movie by 4%. No elaborate and/or expensive standards conversion. I think Compressor will handle this 4% speed up for you. I recently did a post of a technology guide from Quantel and they recommend 24fps as a universal format because it is so easy to transfer across the Atlantic.
And it's important that your software knows that you want to work progressive. You can put 24fps or 30fps in a 60i timeline, but then you've lost all of the technological advantages of working progressive. Your editing software, your computer, your DVD player, your TV will treat it like interlaced video. So as a general rule, pick a frame rate and stick to it throughout post and distribution. Obviously this can sometimes be really problematical--what about archival footage that's 60i? Or you want to work with a shooter and they don't own a camera that shoots progressive. You can convert that to 24fps using Compressor, but it's a real pain. No easy answer for this one.
As or 30 fps, one advantage is that is has more samples per second than 24 fps. 24 fps can look kind of choppy--you especially have to be careful with panning quickly. 30 fps is no walk in the park in this regards, but you have 25% more samples of the image per second than with 24 fps--so definitely less choppy.
In general for cameras, especially with CCD chips, it is easier to create 30 fps than 24 fps. CCD's have been doing 60i for so long that a lot of cameras out there are still based around that technology. Namely, the Sony HVR-Z1, which Link owns. The Z1 doesn't do real progressive imaging, but with the Cineframe modes you can get video that has 24, 25, or 30 images per second. Basically the camera blends the interlaced video to get progressive-looking video. Apparently, the Cineframe 24 with the Sony Z1 is supposed to be really crummy. The Cineframe 30 is supposed to be better, as is the Cineframe 25. Basically the camera is capturing video in an interlaced manner 60 time per second and has to figure out how to make progressive frames with it. With all Cineframe modes, this means tossing out some resolution. The Canon XL-H1 also does a similar process to get progressive frame rates. But going to 30 fps from 60i is easier--basically is the image is moving than you throw about every other field. And the Z1 does Cineframe 30 reasonably but fails with Cineframe 24 (Caveat: this is definitely true for HD, but if you are downconverting to SD it might be OK.)
One option I saw discussed over at a great post and discusion, which talks about frame rates, at HD for Indies is to shoot with the Z1 Cineframe 25 and then slow it down to get 24 fps. I think the easiest way to do this would be with Compressor, but these Compressor workflows are a hassle. However, if quality is important than this is an option.
By the way, Sony is moving from CCD's to CMOS, and the CMOS sensors are much more versatile with frame rates. So the HVR-V1 chips natively captures at 24 fps, as does the new EX-1, the HVR-Z7 and HVR-S270.
One problem is that even if the camera can capture images at 24 fps (or 30 fps), the camera often actually records in video that is actually 60i. With 24 fps that is recorded in a signal of 60i, the camera adds pulldown (see the definition in this Quantel post I did). In post you need to remove the pulldown, otherwise your editing software just treats it like 60i and you lose all the technological advantages of working with progressive video. So to post at 24 fps, you may have to convert the footage to remove the pulldown. This can make post-production very complex. FCP has added a setup for version 6.0.2, which is supposed to automatically remove the pulldown from the V1's 24 fps footage. Newer gear from Sony like the Z7 and S270 are doing a native framerate of 24fps recording in HDV. Tapeless cameras like the Panasonic HVX-200 and the Sony PMW EX-1 handle progressive footage in a much more straight-forward manner.
One drawback with 30 fps is that it doesn't convert well to 24 fps or 25 fps. 60i is easier to convert to 25 fps because it has a lot more samples across time to work with. Basically when you are doing a major standards conversion, a computer needs to generate frames that don't exist by analyzing the image information that is available. The more samples of images, the better that this process can be handled. In short if you have to make 25 fps video or 50i video, than 60 images per second is much better than 30 images per second. (Remember that going from 24 fps to 25 fps can be done in a much easier way by simply speeding up the video by 4%.)
UPDATE
I thought I should add a caveat about the above comments on standards conversion. The workflow that I desribe for speeding up 24 fps to make it a European-friendly 25 fps is a workflow that I've only heard discussed as a step done in a computer, e.g., Final Cut Pro or Compressor. So you could convert the footage on your Mac--but you'd still have to make a master on a 25 fps or 50 fps format, usually tape like DigiBeta or if HD, HDCam or DVCProHD. The traditional way to do standards conversion is that folks just make an NTSC master of their program and then send it to a high-end dub house to get a PAL copy done. This is obviously easier than having to make a master on your own for formats designed to work in Europe. I don't know of a dub scenario that exists for taking a 24 fps master and speeding it up during a transfer to make it 25 fps. If you want a standards conversion done during a dub, then 24 fps may not be any better than 30 fps.
In short, there is no easy answer to this question. First you got to know your camera, know your post options, and know your distribution plans. From there, it's a matter of trade-offs, for quality, time, and money.
24 fps--what's used on film
30 fps--half as many samples per second as regular TV (e.g., 60 interlaced)
25 fps--half the frame rate of regular TV in Europe (e.g., 50 fps)
50 fps--the samples of images per second used in European TV
60 fps--the samples of images per second used in North America TV ( which is usually interlaced)
So people generally prefer the slower frame rates in documentaries and dramatic material because it reminds people of film and the cinema--in short it looks classy. The faster frame rates look like TV and are a must have for sports where things are moving fast.
This post is primarily concerned with the technological differences (and not aesthetic) between different frame rates. (If you want to brush up on your basic knowledge of interlaced versus progressive, check out their definitions from this glossary from Quantel.) So let me outline a couple of the technical advantages of progressive video versus interlaced video:
1) Monitors are rapidly turning progressive--LCD's (both TV's and computers) and plasmas. These monitors can't natively display interlaced video. So some processing has to be done to de-interlace, and very often you end of losing resolution (and lots of it) when the de-interlacing happens. It depends partly on how good the hardware is that does the de-interlacing, but this happens at the consumer's end--so you never quite know how it will look. In short, a cheap TV will do a bad job of this.
2) Better effects work on progressive footage. Lots of effects, including scaling, require the video to be de-interlaced. Final Cut Pro does a lousy job of de-interlacing. If your footage is progressive (and FCP knows it), then it won't de-interlace to do simple things like scaling up the video. This past week, I had to include some letterboxed standard-def video in a high-def show. It looked super soft in FCP. So I used Compressor to do a very slow but very good de-interlacing of the footage, which I then scaled up (but again I didn't even scale this footage in FCP). It looked a lot better with the de-interlacing that Compressor does than the de-interlacing that FCP does.
3) Progressive distribution of video. Computer monitors were never interlaced--so anything for distribution on computers benefits from being progressive. Also DVD's will handle 24 fps. This is great if the video is going to a progressive display, e.g., LCD or Plasma. As mentioned above, interlaced footage needs to de-interlaced to be scaled, but progressive doesn't. So a 24fps DVD will usually look much better on a high-def display than a comparable 60i video.
Important note: DVD doesn't have a standard for 30 fps--it's either 60i or 24 fps. So your DVD player and your TV treat 30 fps just like 60i--if it gets scaled up then you could lose resolution (but not always--again it depends on the display.)
Now let me move on to outline some advantages and drawbacks of different framerates:
24 fps: So if you're going to make a film print, then you should shoot in this framerate. OK .001 per cent of anything shot by professionals will ever end up on film. But because 24fps has been very hot for many years (essentially the Panasonic DVX-100 set that off) there are a lot of post workflows for working with 24 fps. 30 fps is gaining ground though--there are now lots of setups in FCP for it.
24 fps is very close to 25 fps. So a general option for converting for a European master is to just speed up the movie by 4%. No elaborate and/or expensive standards conversion. I think Compressor will handle this 4% speed up for you. I recently did a post of a technology guide from Quantel and they recommend 24fps as a universal format because it is so easy to transfer across the Atlantic.
And it's important that your software knows that you want to work progressive. You can put 24fps or 30fps in a 60i timeline, but then you've lost all of the technological advantages of working progressive. Your editing software, your computer, your DVD player, your TV will treat it like interlaced video. So as a general rule, pick a frame rate and stick to it throughout post and distribution. Obviously this can sometimes be really problematical--what about archival footage that's 60i? Or you want to work with a shooter and they don't own a camera that shoots progressive. You can convert that to 24fps using Compressor, but it's a real pain. No easy answer for this one.
As or 30 fps, one advantage is that is has more samples per second than 24 fps. 24 fps can look kind of choppy--you especially have to be careful with panning quickly. 30 fps is no walk in the park in this regards, but you have 25% more samples of the image per second than with 24 fps--so definitely less choppy.
In general for cameras, especially with CCD chips, it is easier to create 30 fps than 24 fps. CCD's have been doing 60i for so long that a lot of cameras out there are still based around that technology. Namely, the Sony HVR-Z1, which Link owns. The Z1 doesn't do real progressive imaging, but with the Cineframe modes you can get video that has 24, 25, or 30 images per second. Basically the camera blends the interlaced video to get progressive-looking video. Apparently, the Cineframe 24 with the Sony Z1 is supposed to be really crummy. The Cineframe 30 is supposed to be better, as is the Cineframe 25. Basically the camera is capturing video in an interlaced manner 60 time per second and has to figure out how to make progressive frames with it. With all Cineframe modes, this means tossing out some resolution. The Canon XL-H1 also does a similar process to get progressive frame rates. But going to 30 fps from 60i is easier--basically is the image is moving than you throw about every other field. And the Z1 does Cineframe 30 reasonably but fails with Cineframe 24 (Caveat: this is definitely true for HD, but if you are downconverting to SD it might be OK.)
One option I saw discussed over at a great post and discusion, which talks about frame rates, at HD for Indies is to shoot with the Z1 Cineframe 25 and then slow it down to get 24 fps. I think the easiest way to do this would be with Compressor, but these Compressor workflows are a hassle. However, if quality is important than this is an option.
By the way, Sony is moving from CCD's to CMOS, and the CMOS sensors are much more versatile with frame rates. So the HVR-V1 chips natively captures at 24 fps, as does the new EX-1, the HVR-Z7 and HVR-S270.
One problem is that even if the camera can capture images at 24 fps (or 30 fps), the camera often actually records in video that is actually 60i. With 24 fps that is recorded in a signal of 60i, the camera adds pulldown (see the definition in this Quantel post I did). In post you need to remove the pulldown, otherwise your editing software just treats it like 60i and you lose all the technological advantages of working with progressive video. So to post at 24 fps, you may have to convert the footage to remove the pulldown. This can make post-production very complex. FCP has added a setup for version 6.0.2, which is supposed to automatically remove the pulldown from the V1's 24 fps footage. Newer gear from Sony like the Z7 and S270 are doing a native framerate of 24fps recording in HDV. Tapeless cameras like the Panasonic HVX-200 and the Sony PMW EX-1 handle progressive footage in a much more straight-forward manner.
One drawback with 30 fps is that it doesn't convert well to 24 fps or 25 fps. 60i is easier to convert to 25 fps because it has a lot more samples across time to work with. Basically when you are doing a major standards conversion, a computer needs to generate frames that don't exist by analyzing the image information that is available. The more samples of images, the better that this process can be handled. In short if you have to make 25 fps video or 50i video, than 60 images per second is much better than 30 images per second. (Remember that going from 24 fps to 25 fps can be done in a much easier way by simply speeding up the video by 4%.)
UPDATE
I thought I should add a caveat about the above comments on standards conversion. The workflow that I desribe for speeding up 24 fps to make it a European-friendly 25 fps is a workflow that I've only heard discussed as a step done in a computer, e.g., Final Cut Pro or Compressor. So you could convert the footage on your Mac--but you'd still have to make a master on a 25 fps or 50 fps format, usually tape like DigiBeta or if HD, HDCam or DVCProHD. The traditional way to do standards conversion is that folks just make an NTSC master of their program and then send it to a high-end dub house to get a PAL copy done. This is obviously easier than having to make a master on your own for formats designed to work in Europe. I don't know of a dub scenario that exists for taking a 24 fps master and speeding it up during a transfer to make it 25 fps. If you want a standards conversion done during a dub, then 24 fps may not be any better than 30 fps.
In short, there is no easy answer to this question. First you got to know your camera, know your post options, and know your distribution plans. From there, it's a matter of trade-offs, for quality, time, and money.
Subscribe to:
Posts (Atom)