CQ-VBR: Science, or guesswork?

[homerpez]

Anyone have a good way of telling what MAX bitrate and what "quality" setting you're able to use to match a specific file limit?

I'm guessing you can theoretically use the calculated "AVG" on this site's calculator as the MAX, and select "100" for the quality setting. But for what I want, "800" AVG, I think this might be too low for CQ mode. Movie is 112 minutes long.

Any thoughts?

[Scumbag]

I've done heaps of testing on this topic and come to the following conclusions...

There is only one good way to generate to a specific filesize - use 2 pass VBR. Period.

The other way is to use constant bitrate. What a waste of space.

The way to make this process only moderately painful time wise, is to use high quality motion detection vs highest. The difference isn't worth it.

At CQ's above 75, the bitrate used will almost always be the max you have set - typically 2520 kbps. Only really still scenes require less than this to achieve a "quality" rating of "75" according to TMPGEncs archane formula for determining same

Once you hit 80 + you'll find that a CQ max 2520 and a constant bitrate 2520 will basically produce the same file.

Anyway, that's my conclusion after hundreds of test encodes and viewing of resultant streams with a bitrate viewer (and my own eyes)

Others will surely swear that the result was better with a CQ of 100. (I'd be dubious...)

regards,

jr

[adam]

I agree totally. I have seen countless experiments done on this and have done several myself. I would say that the quality level tops off even lower, at around 65.

Of course quality is relative and highly subjective but from every test I've seen the consensus is that encoding at anything higher than 65 only yields a %1-%2 increase in quality over a 65 setting. And depending how high you go the size can increase dramatically.

But to answer the original question...no I really dont think there is anyway to get a half-way accurate estimation. Depending on the source the resulting filesize can very dramatically at the same settings. The good thing is that you can probably rip the movie 3 times in the amount of time you could rip it once with 2-passes, but I choose to do it right the first time.

[SatStorm]

So, if you have a very good source (DVD rip, Satellite rip), then using CQ 65 with high quality (slow), you have only 2% lower quality than CQ75 and highest quality (Very slow)? And in a smaller file?
That is interesting....

High quality and 65 gets much less time than Higher quality and 75. The file is a bit smaller and using "soft" fuction at about 25, you can eleminate those blocks...

Let's start testing!!!!!!


<font size=-1>[ This Message was edited by: SatStorm on 2001-07-06 02:59:47 ]</font>

[adam]

Well I dont know if I'd say that exactly. Factoring in the high quality vs highest quality will make a difference just not a huge difference.

But its true, both of these encoding settings (high CQ and highest quality) are synonomous. They both yeild minimal quality increases and come with pretty substantial drawbacks.

[Scumbag]

>>I would say that the quality level tops off even lower, at around 65.
>>

I was being generous... I agree.

Why else would Hiroyuki Hori have his CQ SVCD template set at a default of 65. I mean - he wouldn't know would he

jr

[homerpez]

OK, now to take this to a new area...

Which encoding method is better to remove stuttering MOTION at low bitrates? CQ_VBR? Or, 2-Pass?

My experience has been this. I could encode 2-pass movies with an AVG bitrate of like 950-1050, and when I use a HIGH MAX rate (like 2200), there's no more stutters in motion. Usually the high MAX compensates for the motion taken. I try this with CQ_VBR, and I'm forced to use 950 as my MAX bitrate, thus limiting heaving this compensation.

Now, my current problem. I'm giving up on it, but so I know for future movies:

I tried to do "Falling Down". It's 112 minutes, so I had to lower my AVG to about 827, actually I used 850. There was constant scenes I could find that had motion stutters on them. And this time it's not because of my MAX bitrate, that's 2200 again! Should be high enough. What's more, there were on LOW MOTION SCENES!!! You'd think it'd be when something explodes, but this is when people were in a chair turning their heads...

Which should I use? Or, is an AVG of 850 just too low?

[thebach]

I know that about 1 out 0f 12 DVD's I do have motion problems when going to SVCD, even when I use the same settings as the rest of them. I don't know why this is, it must be something to do with the DVD, because higher bitrates don't even help

[homerpez]

You know, thebach, that answer is very comforting to me right about now.

No one ever gives that answer, and that's what I was suspecting.

Though I don't think it's the DVD exactly, maybe just the length of the movie itself...? Maybe with a longer movie it has trouble with calculating an AVG, so it messes up in spots?... I dunno at this point.

I agree with you. I'll just let 1 of every 12 be "watchable" instead of "perfect".