MPEG-1 Multi-Pass VBR Problem

[Hacksaw]

Peoples,

Looking to see if anyone else has been having this issue with CCE 2.5 and knows of a fix, or if I should just give up.

I am on my 10th attempt or so at encoding an XVCD using CCE's multi-pass VBR MPEG-1 support. I have been trying 4 passes, with the min = 1150, max = 2500, max average = 2400 and started with image quality priority set to 24. I was amazed at how small a 43 minute tv show could be encoded when compared to TMPGEnc's VBR setting. However, I was getting a bit pixelation/blocks on some scene transitions. Reducing the image quality priority increased the problem, and although increasing the this value reduced the number of occurences, my last attempt with it set to 80 still produced a few scenes with the problem.

I noticed newer version of CCE does not include MPEG-1 VBR support. Was this causing problems? Any clue if this is a known issue?

Thanks,
Hacksaw.

[jankster]

I have never done mpeg1, and never have the bitrate below 3000 avg. At this super low bitrate you should be using "Ultra low birate" in the Quantization matrices"

[bb]

Another good advice would be to use TMPGEnc instead of CCE, because CCE's MPEG-1 quality is known to be inferior.

bb

[Hacksaw]

Thanks for the information. I have found that beyond the issues I outlined above, CCE provides a fairly good mpeg-1 encoding. I did a VBR mpeg-1 using TMPGEnc but had it turn out almost 200 meg larger than the equivalent CCE encoding with no noticable difference in quality. I think this has to do with the fact CCE support max, avg, and min, whereas TMPGEnc only supports max and min.

John.

[waldok]

Did you try CAnopus Procoder for your mpeg1 encoding ?

Waldok:cool:

[segfault]

I think this has to do with the fact CCE support max, avg, and min, whereas TMPGEnc only supports max and min.
You will get far better results with TMPGEnc using CQ mode than regular 2-pass VBR. Try it again with CQ and see what you get. :D

[Hacksaw]

But can you predict the CQ file size? I thought that was too hard given all you have is the quality slider and not a max bitrate?

[segfault]

It's pretty easy, actually. There is a forum devoted to the subject over at kvcd.net. Basically, you just use SansGrip's sampler filter in your AVISynth script to take a sample equal to the length of your GOP once every minute, then encode it and plug the file size into a formula.

If your GOP is 18 frames, you would put sampler(length=18) in your script and encode at the CQ of your choice. Then you do
((total frames/total minutes)/GOP length)*sample file size

I have never done it with a GOP length other than Kwag's 24, but I think it should work in principle. Have a look at Kwag's forum for more info. It saves a lot of time.