Mastering Formats for different purposes

[Alan Van Vliet]

I am in the process of prepping a large number of very short clips for a tennis instructional web site, using predominantly 8 – 12 second clips. Sharp, stutter free video and clear sound quality are my highest priorities.

I am filming with a XL1S at 1/60th of a second in normal mode. I am using Windows 2000 Pro, Premiere 6.02, and will be using Quick Time Pro for encoding for the present time. (I plan to go to a Flash format as I am authoring in Dreamweaver) either with Sorenson or Cleaner).

My basic workflow is as follows:

1) Save one master (uncompressed AVI file of the clips for future use) for future conversion to Flash format and /or interactive DVD).

2) Save another master to be prepped for Web use.

3) Use Quick Time Pro to encode.

My questions.

1) What format is best for the original master? Interlaced? If so, what type of settings accomplishes this?

2) What format is the best for importing to Quick Time Pro?

3) Which of the current Quick Time Pro codecs would be the best for this situation?

I have been searching the archives and was unclear on the different "formats" for interlaced and de-interlaced, the latter being suggested for web preparation.

Thanks in advance for any thoughts.

Al

[Rob Lohman]

If you are shooting in interlaced (what normal mode does on the
XL1S) then I would keep it interlaced throughout your pipeline,
including the final exports and mastering formats.

QuickTime (Pro) basically has two formats that are mainly of
interest. These are Sorenson (3) and MPEG-4. With the former
you may want/need to purchase the full version so you have
much more control over the compressor.

Keep in mind that such (QuickTime) encoding is usually done for
the internet and most people half the resolution of the file (which
for NTSC: 720 x 480 (0.9 pixel aspect) -> 640 x 480 (1.0 pixel
aspect) -> 320 x 240 (50% resolution)) and sometimes drop the
framerate to something like 15 fps to make sure the file is not
too large for the internet.

If you drop your resolution in half you are basically dropping one
field of your interlaced signal, so the result will be a progressive
(not really de-interlaced) signal instead of interlaced (which is good
in this case, ie, internet delivery).

Keep in mind that today the Windows Media format is also used
a lot for internet delivery and I think it would be wise to offer both
formats so people can choose what they want to use
(not everybody can play both as well).

You can use Premiere to encode to QuickTime, you do not need
to do this in QuickTime Pro itself (not even sure that is possible
on a PC). If you really want to do the encoding outside of the
computer go with an Uncompressed format to that encoder if you
can.

Premiere 6.02 is pretty old (these days), so I'm not to sure how
everything works there (I'm pretty sure that version cannot do
a Windows Media export). Did that version already support native
DV editing?

[Alan Van Vliet]

Rob,

Thanks for your time and excellent thoughts.

I am kinda stuck with Premiere 6.02 for a while, and as it is getting old, I thought the compressors in QT Pro would be more up to date versions than Premiere's. (The QT Pro does work in a Windows based enviroment).

BTW, the 6.02 Premiere does allow you to export to Windows Media in a variety of formats.

I was planning on using a 320 X 240 frame for viewing. However, I did not understand the reasoning until you explained the %'s. So by halfing the frame size, it works better with the interlaced format. (Same with the frame rates, I would assume).

As my subject is constantly moving, I assume that I will need a fairly low keyframe rate. Is there a similar relationship (halving) in determining this factor?

I also take it from your comments that I should be using the .9 Pixel Aspect ratio instead of the Square Pixels settings in the Project Settings for the 720 X 480 size. I have never really understood how this worked before.

Thanks again for your comments.

Regards, Al

[Dan Euritt]

you will want to let the encoding app, not the editing app, convert it to square pixels, because that is what is used on computers.

the best results by far come with two-pass encoding, and i don't believe that you can do two-pass encoding with qt pro... so you will have to spend big bucks on squeeze or something similar to get decent video made in the qt format... but squeeze is a pretty cool app, and you can also create two-pass flash and some killer h.264 mpeg4, which will be really important later this year, when the h.264 software players finally start coming out... you may look at the quality of h.264, and completely bail out on the idea of using the current implementation of flash.

as an aside, if you have the right version of nero, you can also create h.264 stuff.

the stand-alone windows media 9 encoder is a free download from microsoft, check it out to see if it'll do two-pass encoding... your old version of premiere may not allow for two-pass encoding and the full flexibility of the wmv encoder... i would export a native dv avi from premiere instead... the best pq with wmv comes with using the full 640x480 frame size, and the full 30fps rate as well... can you use mono audio?

computer monitors are not interlaced, so that is not an option... high-end windows media files can be encoded in their native interlaced mode, ala hi-def video, but i don't believe that is intended for computer viewing.

[Rob Lohman]

The reason to lower resolution and/or framerate is to lower the
amount of data the encoder needs to handle and thus you can
create a smaller sized file which is important for internet delivery.

Broadband internet connections are becoming much much more
common, but if you make a movie of let's say 50 MB (which is
easily reached!) and 50 people download it of your server you
are already using 2.5 GB of bandwidth, which your provider might
not like very much for example (or you if you need to pay extra
per MB for example).

Internet video should be in square pixel (= 1.0) aspect ratio, this
means that a pixel is as high as it is tall (ie, square). Normal (NTSC)
DV has a pixel aspect of 0.9, this means that the pixel is taller
than it is wide. To convert this to square pixels multiply the
horizontal resolution by the pixel aspect ratio:

720 * 0.9 = 648 (which is commonly used as 640).

So normal full size output in NTSC DV at 1.0 PA would be 640 x 480.

To make it easier on the encoder (and get the best quality) you
want to make sure the final output resolution is an even based
number that (if possible) is a percentage number without fractions
as well.

So usually people use 50% which yields 320 x 240. Some people
might even go as low as 25% which yields 160 x 120.

Keep in mind that if your footage is in 16:9 or you want to add
letterboxing (black bars) it is much wiser to crop these for a web
encode so you are not wasting valuable bytes and bandwidth on
something like a black bar.

In this case the actual footage window might be 640 x 360. So
120 pixel (lines) are used for the black bars, 60 on top, 60 on the
bottom.

It would be far better to output this at 320 x 160 (make sure you
crop and not resize, ie, the black bars ARE NOT outputted) than
at your normal resolution of 320 x 240. Saves you 80 lines of
nothing to encode.

Hope this gives you some ideas. In the end it all boils down to
experience and trial and error. Both in what works best, quality
versus filesize etc.

[Alan Van Vliet]

Guys,

Thanks for your valuable input.

Dan, I do need stereo sound. The focus of my teaching is based on the sound of contact, so quality sound is a very important part of my presentation.

I am currently editing 48000 Hz uncompressed stereo sound with a 1/2 second interleave and using an IMA 4:1 compression, resulting an a fairly large audio file.

In reference to Rob's comments, does the keyframe rate coincide with the same "halving concept" as with the resolution and frame rate?

I am anticipating using a fairly low keyframe rate as my subject will be in constant motion, although filmed against a black windscreen background.

Once again, this is for very short clips of 8 - 12 seconds in duration.

Thanks again for your thoughts.

Regards, AL

[Rob Lohman]

Keyframe rates are typical with inter-frame compression methods,
this indicates when it will take a full frame as a starting point again
instead of a previous frame. When this number is high it will be
futher compressed but usually results in a softer and sometimes
corrupt image etc.

[Alan Van Vliet]

Hello Rob,

Thanks for your response.

If i understand your comments correctly, the compressor initially determines this rate by analyzing changes in sequences of frames.

If I want to tweak this rate to improve upon quality (at the expense of file size) is there a general range to use depending on your needs for quality and on the amount of motion and/or change in your video?

Should this keyframe rate be proportional to the frame rate (1:2, 1:1, 2:1 etc)?

Regards, AL

[Rob Lohman]

The best way to answer that is through trial and error. It depends
a lot on the codec you are using and the footage you feed it. It is
always good to use a keyframe on a frame that looks drastically
different than the frame before (ie, a straight cut for example).

Some (usually bought) codecs/encoders allow you to set keyframes
to that level. However you may not need such detailed control
(takes a lot of your time as well), or some codecs/encoders may
support to such detection themselves. In the end the only way
to see what looks good enough at is small enough is to test, test
and do some more tests.

Sorry....

[Alan Van Vliet]

Thanks again for all of your imput. I have been running a series of tests on a 12 second clip. I will post results when completed.

A question regarding formats. When mastering a 12 second clip for multiple formats, the size difference between a Microsoft AVI and a compressed Microsoft DV AVI file is signifigant. (396 meg - 48 meg).

Is the Microsoft DV AVI format suitable (a good master) for different formats, or is it better to stay with the uncompressed file?

When I import both formats into QT Pro, I end up with the exact same file size after using the Sorenson 3 compression. (1.4 meg)

Disk space and file size is a factor as I will be working with hundreds of very short clips.

Any thoughts?

Regards, AL

[Rob Lohman]

Keep in mind that just plain "Microsoft AVI" doesn't tell anything,
that's just the container format. From your post however I conclude
you are using uncompressed AVI (ie, the "codec" is none, or
uncompressed). DV AVI is using the DV codec to compress the
data.

DV is a (visually, usually) lossy codec. This means that it throws
away information to reduce file sizes. This is usually visually
lossless (ie, most people can't see the difference). However, in
the end you are losing information. Whether this is acceptable
depends on things like:

1. whether you have the harddisk space

2. whether you can see the difference

3. what your output format is

If you are only going to release this on the web or on a DVD then
I'd say there is no harm in going out to DV AVI (MAKE SURE that
you select progressive (or none for interlaced) when outputting
to keep your progressive look!!!). For any HD or film output it may
be worth it to keep it in an uncompressed form.

[Rob Yannetta]

I've been experimenting with different file formats myself recently.

I've found that for quality, use Quicktime with the Sorenson 3 video codec and the MPEG4 audio codec. However, these file are LARGE. Excellent for streaming, though. Target audience: Windows, Mac.

Using Quicktime's ability to save to the new MPEG4 video & audio standard is great for streaming, creating smaller files, but the quality lacks a bit at 320x240 with stereo sound. Your target audience, however will not discriminate based upon operating systems: Windows, Mac, Linux, etc.

Microsoft's simple "Movie Maker" in Windows XP does a good job compressing video and audio into a 512Kbit stream. The stereo audio is acceptable and the video is not bad at all. Target audience: Windows only. This is why I choose not to use it.

Finally, there's Real. Unfortunately, Real's content creator crashed every few minutes not allowing me to create any files with it. I have yet to get it to work without bluescreening the entire OS. I believe once they can clean up thier software encoder, they may have a good product because they have players for Windows, Mac and Linux. The Linux player is very stable and high quality.

In conclusion, I sacrifice filesize for quality and let people stream Quicktime using Sorenson 3. However, everyone's different.

I hope this gives you a bit of insight. Good luck!

[Alan Van Vliet]

Thanks guys for your thoughts and great ideas. It has been very helpful for establishing a good frame of reference that I have been able to experiment from.

I am currently using the Sorenson 3 with MPEG audio and the quality is very good. A 12 second straight clip (with no extras) comes in at 1.4 meg though.

As I plan on having a substantial number of clips, this does create bandwith problems as Rob Lohman pointed out.

I plan to try some different compressors (and/or settings) to see if I can get the file size down somewhat without losing too much quality. I'll try the MPEG 4 and the H263 to see how they compare.

I'll post results when done.

Regards, Al

[Alan Van Vliet]

When considering file size and format for web delivery, the following question came to mind.

When the same video clip is viewed on the web multiple times from the same computer (e.g. instructional clips) does the provider incur the same bandwith "charges" for each instance?

Or is it a "single instance charge" so to speak, and can be viewed multiple times or even in multiple sessions?

Any thoughts?

Thanks in advance. AL

[Rob Yannetta]

<<<-- Originally posted by Alan Van Vliet : When the same video clip is viewed on the web multiple times from the same computer (e.g. instructional clips) does the provider incur the same bandwith "charges" for each instance? -->>>

With Quicktime, there is an option called

cache="true"

to cache the video in the user's memory so it can be played ad nauseum without having to download it again.

Here's an example I use in an HTML file:

<object>
<embed src="filename.mov" width="480" height="286" autoplay="true" cache="true" pluginspage="http://www.apple.com/quicktime/download/standalone/"></embed>
</object>

Always add 16 pixels to the height to add the control panel.