[MEncoder-users] new doom9 codec comparission (submission)

Corey Hickey bugfood-ml at fatooh.org
Thu Dec 15 04:29:46 CET 2005

Doom9 Feedback Hotline wrote:
>>> Then by all means be undersized, but beat the competition while
>>> you're doing that ;)
>> At the same bitrate, my best snow encodes do beat my best mpeg-4
>> encodes; however, if snow isn't reaching the target bitrate for
>> you, that could mean it doesn't have a chance to look significantly
>> better.
> Could it be that you're only comparing to lavc?

Well, yes, and strictly within the context of encoding The Matrix for
this comparison.

> I don't know why that
> is, but the Linux folks seem to avoid XviD, whereas the Windows crowd
> gets the better open source codec.. XviD in all my tests didn't leave
> lavc a chance. We'll see how it does this time, but it already got a
> severe beating in the speed comparison.. XviD delivers 92 fps.. now
> that's a number I like (but I have one ASP codec that gets up to 139
> fps).

It's not that I have anything against XviD. Lavc MPEG-4 has been stable
for a longer time, and, back when I tested video encoding more often,
was able to beat XviD in my own tests. That was a couple years ago, and
XviD has been under heavy development since then while relatively little
has happened to lavc MPEG-4 encoding.

Preparing for this comparison has been the first time I've done any
intensive video encoding tests in quite a while. I'm curious to try your
XviD encoding parameters and compare the result to my best lavc MPEG-4

>> Also, Rich Felker just pointed out that snow quantizers aren't
>> really comparable. It may well be that having the same minimum
>> quantizer for both codecs means snow is disproportionately limited
>> in how accurately it can encode a movie. Rich, can you please
>> confirm/refute that and maybe elaborate a bit?
> I don't want to limit anything.. but apparently the rate control
> isn't capable of doing its job properly if there are no limits. That
> reminds me of the early XviD ages where you had to set the minimum
> quantizer to 1 and limit the maximum quantizer as well, or XviD would
> use quantizers that sometimes made no sense at all. That has long
> since been fixed though and today you can use the whole quantizer
> range and rate control picks the proper values. IMHO, rate control
> should not be told which quantizer ranges it can use, but rather it
> should make the proper decision on its own while respecting the
> average bitrate requirements.

Yeah. See Loren's email from early this morning for an explanation.

>> Yes, potentially. In the 2-pass method, though, finding an
>> appropriate vqscale value for the first pass already takes several
>> tries, so that was never really 2-pass either. My approach isn't
>> the panacea of snow encoding, but I'm pretty sure it reduces the
>> amount of guesswork and wasted passes.
> I hope the comparison ends up motivating people to fix what doesn't
> work as it should rather than complaining about a windows loving DVD
> pirating crowd.

Who does that?

> I think this is an excellent opportunity to show to
> the world that open source can compete, even beat many of the
> commercial competition without putting undue restrictions and
> limitations upon the user. XviD is an excellent example of how open
> source beats the commercial competition, and I'm hoping x264 will
> give the competition a good run for the money as well - it has
> advanced by leaps and bounds in the last year (incidentally the year
>  it got a lot of tools that worked under Windows - not that this is
> much of a problem with Snow, there's ffdshow (albeit it doesn't offer
> two pass encoding), and MeGUI, but neither has really catched on so
> far so I guess it's not the tools but the codec itself, all that rate
> control trouble will make people think twice about using Snow).

Nobody should be using snow at this point for anything but testing
purposes. It could change incompatibly in the future. What makes snow
remarkable is that it can compete with mature, highly-tweaked and
-optimized codecs despite not being finished. If Michael ever starts
working on it again, I wouldn't be surprised if it won your comparison
someday. For now, though, I can definitely understand why you have
problems with it.


More information about the MEncoder-users mailing list