Why use motion based curve modulation?

[Zarxrax]

Motion based curve modulation gives scenes with motion less bits than still scenes. I thought the whole point of doing SBC in the first place was so you could spend more bits on the fast motion scenes?? And fast motion scenes will obviously need MORE bitrate than low motion scenes, so this setting completely boggles me. Yet in the guides it tells you to use this! Could anyone please explain this to me??

[Koepi]

You have wrong thoughts there. SBC is there for giving something like constant quality instead of constant bitrate.
The eye can't resolve much details in fast motions scenes. So you can lower the bitrate there.
You are really sure you read the guides and the NOE?
You can give negative values there to do stupid things like spend more bits on fast motion scenes.

You shouldn't feel and behave so wise if you aren't. People use this over a year and you now discover something bogus? Really unlikely.

[Zarxrax]

I am not trying to say that everyone is wrong, just that I didn't really understand it. Lately I've noticed in my encodes that the fast motion scenes look like utter crap compared to the perfect low motion scenes. Even without the motion based curve modulation though, the fast motion scenes are already going to look worse, as they would require more bitrate in order to stay the same quality as low motion scenes. Giving them less bitrate would simply make them look even worse.
SBC is there for giving something like constant quality
Using a positive value for motion based curve modulation sort of defeats that, doesnt it?

By the way, I tried to consult NOE before posting this, though it is not working. I did not know that you can use negative values, and that is very helpful for me to know, thank you.

[Zarxrax]

And of course the very second I post that, NOE comes back up again, after I'd been trying it for almost 24 hours =P

[Zarxrax]

Back to my problem at hand, what would be better to use for giving more bits to the fast motion scenes, motion based curve modulation or curve compression? I realised that the curve modulation helps to keep the predicted filesize accurate, and thats quite important for me.

[Koepi]

If I were you I would use motion based curve modulation for that. It doesn't break size predictability if set in "sensible" ranges.

Try something like -10 first and see if the results work out good for you. You could use asymmetric curve compression as well, using something like 25 for low and 15 for higher than average.

BUT since DivX3 is rather outdated/old now, you would always have a problem:

doing that suggested stuff above you'll loose quality on low motion scenes where it is really obvious. That's why we tried to move it to the action scenes....

Best regards,
Koepi

P.S.: tripod/lycos had a downtime since 8.2. because they restructured themselfes, I just had the possibility to log in again a few minutes ago and recreated the site from my backups (it's a new URL now...). Dark Avenger (h3adache) suffers from the same problem, i should PM him that he should recreate his site too...

[OUTPinged_]

zarxrax, you are missing something.

>Motion based curve modulation gives scenes with motion less bits >than still scenes

Positive MBCurveM values are decreasing bits for _frames_ with higher motion values. It has nothing to do with "scenes".

The idea behind that is that it is usually a bit harder to notice lower drf at frames with high motion values.

The bad thing is:

1. Motion detection code isnt perfect.

2. there are some high motion scenes where excessive blockyness appears even at drf3.

3. if you have half of your frame with very high motion and other half perfectly still, all frame will be affected with bitrate change.

So, just dont use it with high (20+) values. That feature works best on content, where you need to have smooth panning at cost of hi-mo frame blockyness.

ie, 20%m.b.curve.m and 20%symm.curve.m will give you 36% decrease in bits for very complex high motion frame. So usually it will come around 3 drf levels lower than your average drf. So it can reach 7-8th drf easy, and that equals blocking. Easy :) Cap drf man. That will screw filesize prediction tho :P

Other way is to use motion based cripsness modulation. From my experience when you use both crispness and curve modulation, the encode comes out bad-looking more often.

[poopity poop]

zarxrax, you're the man.

Finally someone else who enjoys higher quality high motion scenes. I've stopped using both of those settings a while ago(set them both to 0). The low motion scenes look fien and the high motion scenes look great.

I unerstand the point of them: The high motion scenese get more bits, but not unesseary amount of extra bits...my personal preference is that...who cares. the low motion scenes look great, the high motion scenes look better, its a matter of personal preference.

In fact: everything is a matter of preference with some things(not many) with certain standards.

[OUTPinged_]

poop, sometimes bitrate jumps up to 6-7mbits on some 1h/cd highres encode's scenes with mbcm=0. I personally use at least 10% of mbcm in anime content (*Cough* kenshin ovas *cough*). Well there are extreme cases like Lain (25% of assym curve compression up, 35% of mbcm, drf caps at 5 on still scenes, at 6 at motion intensive scenes).

You use mbcm=you have to use at least drf cap to not get your encode too blocky, that is my point.

Well it looks like i have to edit alot of scenes manually anyway, so it doesnt matter after all...

I remember one scene in Lain eps where bitrate averaged at 2.3mbps at drf12 =) You cant go away with automatic encoding in these cases.