VBR vs CBR = Multipass vs Singlepass

[Edwin Street]

I always seem a little confused when I come to these options. I'm usually using MPEG Streamclip or Compressor.

I've found that Variable Bit Rate (VBR) isn't as good quality as Constant Bit Rate (CBR). Some people say that VBR takes longer because it does two passes but gives better quality. But I want to get the story clear because when I play back the video of a VBR the bits per second alternate from as low as 1Mbps to 9Mbps with burning a DVD. But if I burn at CBR it stays on 9Mbps the whole time so wouldn't this prove that CBR gives better quality.

When encoding a video in MPEG Streamclip, I'm not 100% sure but I assume that (when encoding with H.264) multipass = VBR and singlepass = CBR, can you please confirm that this is correct.

If I look at the quality of a video encoded with multipass ticked on or off, I can't tell any difference. The singlepass come out with a slightly higher bit rate. With storage being so cheap these days, I'm honestly not even thinking about file size vs quality.

I want the highest quality video when uploading to youtube and I'm not worried about saving disk space. For example; I uploaded a 30 minute, 1080p, video to youtube that was 12GB in size. It took 3 days but I didn't mind the wait as it's going to be on youtube forever. I encoded in MPEG Streamclip, H.264, quality 100%. The video can be seen here:

Short film "Cult" - YouTube

This was the best settings I could do using MPEG Streamclip and youtube still doesn't do a very good job of encoding.

I'd like to hear what you prefer to use, VBR or CBR, Multipass or Singlepass?

[Les Wilson]

The datarate or bitspersecond of the video is really the thing that influences image quality the most. That is, the more bitspersecond, the more information about a region of the image can be put into the stored video. The more information, the more accurate a representation of the original. Up to a point.

What may be helpful to understand is that for any content, there's a point at which adding datarate won't improve the image quality during playback. When you add that Youtube or Vimeo is going to recompress your file into something that can be downloaded and played over an internet connection, there's a point at which adding more data in the version for upload is overkill and you've wasted the time to generate it and upload it. The download bandwidth forces the maximum data rate.

Storing the original clips or final edit is a different story. Saving the original quality means you have, well, the original and you can never get better than that. So exporting the final edit in the same format as the original clips is like generating the final edit at that highest quality of quality possible. Anything more is overkill or you are uprezzing it for some other reason.

Back to the point, why compress a video at all? Basically the answer is so that you can deliver it in a practical way (over the web, on a DVD) and so that it can be played on a wide variety of computers that aren't as honkering as your edit suite or have special software for the particular CODEC the camera used. H.264 is a standard for doing that. MPEG2 is another.

As you have discovered, you can generate H.264 in ways that make the file smaller or larger. The image quality is pretty much going to follow that the larger the file (data rate), the better the quality... to a point. The size of the playback window affects how it looks. You may generate a 720p file but if watched in a 480 sized window, then viewers may never see the difference between the 720p at 2Kbps versus 5Kbps.

As for VBR vs CBR, the principle is that areas of the video that need more data persecond to keep a good looking image (e.g. a moving camera) can have a little higher data rate because the compression algorithm will undershoot or "borrow" the data rate at a later point so that it averages out... to a point (if there are no sections that data rate can be "borrowed" from then ....)

Multipass is giving the compression algorithm the luxury to spend more time tuning the resulting compressed output so that sections needing more compression time and data rate can benefit from techniques that handle certain things in the first pass and other things in the second pass. ALso, intermediate information about the image can be stored in the first pass and leveraged in subsequent passes and eventually thrown out. No sense in doing that if there's only one pass.

Lastly, algorithms matter. A file generated by Compressor is not identical to one generated by Streamclip even when the same parameters are used. Some software is better than others.

[Ron Evans]

If you are putting more than about 75 mins on a SD DVD then multipass VBR will give a better image than the CBR bit rate needed. Average data rates may look the same but the VBR encode is able to use much higher bit rates when needed and very low if there is no motion/detail. I shoot theatre shows usually around 2 hours and always use VBR 2 pass encode TMPGenc software for the encode for SD DVD.

Ron Evans

[Josh Bass]

My one caveat, as a non-expert in this area, is that I had an issue with a short film I did where every time a scene cut to black (I did this several times for effect), faded into or out of black, or went from titles/credits to the actual movie content and vice versa, most DVD players playing it would hitch and skip for a while before getting back on track.

I eventually tracked this down to using VBR instead of CBR for my encode. Apparently the dramatic change from a pure black screen to full color and motion required so high a bitrate variation that it would throw most players into a frenzy. Once I started using CBR these problems disappeared.

[Ron Evans]

Set max video bit rate to 8K in the encoder then it doesn't happen.

Ron Evans

[Josh Bass]

See, I think I even did that. I even went down to 6K once! Still did it.

[Ron Evans]

What encoder were you using?

Ron Evans

[Josh Bass]

This was years ago, so i was using vegas 6 or so with the native encoder.

[Ron Evans]

That far back the encoder was not very good.

Ron Evans

[Josh Bass]

Ah.

Now I have FCP 6 and compressor from that suite.

I guess I'm still gunshy about using VBR and would rather play it safe with something I KNOW works than mess around with uncertainty. Especially since the problem I had didn't occur on EVERY player, only most.

[Ron Evans]

I understand. But for most of my DVD's they are 2 hours or more and there is a big difference VBR and CBR for me.

Ron Evans

[Josh Bass]

Oh yeah good point. My longest film is 17 minutes. All my works together still equal about an hour!

[Edwin Street]

Whenever I play back one of my videos on playstation3 I find that fade to and from black the bitrate always jumps up pretty high.

Having said that, is there software I can download that will show me the bitrate of a video every second. I've found that quicktime, streamclip and vlc only show the average bit rate and not the current bit rate. This would be very helpful as I always have to burn a video to DVD and play it back on the ps3 to see the current bit rate.

[Geoffrey Cox]

I use Compressor and also only master short films (35 minutes max so far) and find 7 Mbps at CBR gives the best results by far. If one has the disc space then CBR is a no-brainer to me. Otherwise, VBR makes sense as to use CBR one would have to set the data rate very low to get it all on.

[Josh Bass]

If I run into space consideration issues in the future, I may go back to trying VBR but will likely stick with CBR in any other case.