Anybody know about VBR and CBR?

Question

Hi,
I'm using PPro CS3 at the moment since I can't get CS4 to work very well on my computer.
On exporting.
I would like to export- adobe media encoder -to Flash video and to quicktime.
I'm sure I used to be able to set VBR 2 pass etc etc with different mbps.
Have I lost these? Can you point me in the right direction?
I only seem to have CBR settings?
I am using CS3 v 3.0 (I don't want to install update to 3.2 as when I did this all audio sync went out by about twelve frames- the ability to batch capture vanished too)
Thanks
scott mackenzie

Answer 1

Scott,
That particular video was encoded using VP6 2-pass VBR, 650kbps, 480x352, from Sorenson Squeeze for Flash v5.0.4.10. I use the VP6-E profile--the VP6-S profile is more if you're exporting HD-sized videos, but since this is originated from SD there's no need to incur extra decoding cycle. I turn on auto keyframes, and generally leave it around the Squeeze default; you may need to edit this setting depending on the footage you're encoding. Leave compression speed set to "best"--this is a counter-intuitive setting that actually should be labeled "compression quality" or have options ranging from "slow" to "faster". "Best" means the encoding runs slower, but results in a higher quality output. Strange, but true.
I crank the minimum quality slider up to about 70-75; this keeps the compressor from getting a little too overzealous. I find that it prevents some blockiness when transitioning from one scene to the next--and in my experience, the added file weight is pretty minimal for the quality gained. If you have fairly static footage with few transitions, you can gain better compression ratios by leaving the minimum quality slider set to a lower threshold. It's these settings that let VBR do its thing.
I never, or at least I haven't yet, enabled the "drop frames" setting. Unless you're trying to compress and distribute HD-sized videos, I think the prevalence and speed of high-bandwidth Internet connections negates this. On the other hand, if you need to stay within a certain bitrate threshold, like for some sort of hardware playout device, you'll probably have to experiment with this in order to produce files that meet the requirements of the hardware. For general web distribution, I've never found the need for this.
For sharpness and noise pre-processing, you'll again have to experiment. There's no magic number for these: it's going to depend on your footage and your tastes. They will both slow down encoding, and can have either a positive or negative effect on your video. I will sometimes use noise pre-processing if I'm encoding from DV or the like, to try to eliminate some of the noise and grain. This makes for a more efficient encode, because the compressor isn't trying to preserve details you don't want to see. These two settings sort of need to work in concert with each other, because one is going to blur the footage to remove noise, and the other is going to sharpen it and bring it back to a certain extent. I suggest experimenting with a short section of video, going with the extremes and then a couple settings in the middle--you'll quickly see the difference. I probably had sharpness set to 5-7, and noise pre-processing set to about 1-2, for this particular video. I think the result is quite clean.
Finally, I leave the minimum/maximum VBR and undershoot settings where they are--haven't needed to tweak these yet. These get pretty technical, and are again really the crux of VBR 2-pass encoding. They are what allow the compressor to "bit budget" an allocated bitrate for a given passage of video, and make sure that the final output stays as close as possible to that allocated bitrate. So far, the defaults have done the trick for me.
One note that I should probably explain, and that's to do with frame size. You might notice above that my output dimensions are 480x352, and not 480x360, as you might expect for a 4:3 video. This is because On2 VP6 (and many other codecs, like H.264) operate optimally in what is called mod16. What this means is that a mod16 codec will first break an input stream into 16x16 blocks, then 8x8 blocks, and then other sizes like 8x4 and 4x4. If you use a dimension that is evenly divisible by 16 (like 352 pixels high) you maximize the efficiency of the codec. I suggest doing this by cropping, and then scaling--not by scaling alone. I typically leave the "Maintain Aspect Ratio" radio button selected in the options, set one dimension (height or width, whichever), and then experiment with the crop filters until you get the other dimension evenly divisible by 16. Once you find the magic number, save the crop filter as a preset and save yourself some headache later! The crop filter will have to change depending on your export size--sometimes you'll crop vertically, and sometimes you'll crop horizontally. And it will also change if you're exporting to a 16x9 file. Sadly, that math stuff we all got into video to avoid will rear its ugly head yet again ;)
Good luck, and let me know if you need any other assistance. The biggest thing to remember is that there is NO magic formula to encoding, regardless of the source footage, the destination, or the codec used. Experimentation is really the only way.