Divx5.01 Psychovisuals, anyone try it yet?

[Nazgul]

According to DivxNetworks, they've made some improvements to the Psychovisual Enhancements. Has anyone tried it out yet?



Nazgul

[TheXung]

There may be improvements, but under certain conditions, I still get horrible errors. However I have only checked it under the strongest setting.

[DJ Bobo]

I tried the psycho light setting with quality based 100%, I havn't got any size improvement, so I won't use it.

[tripnotik]

@bobotns:

Of course you don't have any size improvement. Psychovisual enhancements are not there to reduce size, but to try to allocate the bits in a more optimized way. And it does so on a frame and on a macroblock basis. So by setting quality at 100%, it was only operating on a macroblock basis, and since the quality is very good already at 100%, I doubt that you would see a difference.

[Dali Lama]

I agree with tripnotik.

I have tried the new Psycho Stuff and its pretty darn good on light with anime. That's what I've tried so far. It produces a better looking encode personally.

Dali

[Dali Lama]

I actually have a question about Pyscho too. Is it to be used for the first and second pass? Or are you supposed to use it only on the second pass, like Xvid?

I'll check out divx.com for answers,

Dali

[Doom9]

you use it on both passes.. during the first pass additional info is written to the log which is used to the 2nd pass... so if you turn it on in the first pass and then make the 2nd pass without it you get horrible results (I did my 2nd encode like that and went complaining to DXn about it).

[unplugged]

:D:D everyone has tried with light or strongest setting :confused: :D:D

Well, I have tried with *normal* setting and I obtained a little more but good detail with some Matrix content (720x416 cropped at 1600 kbit/s)

NO ARTIFACTS :cool:
------------

[hulkenstrong]

I used 5.01 and strong physial enviroment on Matrix at 354 as bitrate and the final movie quality is really good. So I´m gonna keep testing/using physical enviroment.

[Acaila]

Dali Lama wrote:
Or are you supposed to use it only on the second pass, like Xvid?
With XviD you should also use it on both passed to get a better result, otherwise the bits won't be saved because the underflow protection kicks in.

-h recently commented on this in the XviD section.

[Dali Lama]

That's interesting Acailia, but how is one to use lumi masking on the first pass if it is "greyed out" in the settings tab? Even more interesting is the fact that I read the Xvid forum everyday and have not seen -h mention that.

Thanks,

Dali

[Doom9]

the changes haven't been made it into the frontend but I recall seeing that post from a developer hinting that lumi masking would once again be a two pass thing..

[Teegedeck]

Originally posted by Doom9
the changes haven't been made it into the frontend but I recall seeing that post from a developer hinting that lumi masking would once again be a two pass thing..

Hello Doom9,

I couldn't find a statement like this. Wasn't it only greyed out to PREVENT users from activating lumi-masking in 1st pass by mistake?

[Acaila]

Originally yes, but now they're of the opinion that using it for both passes would actually be better.

When it was first made unavailable in the 1st pass it still had quite a few problems as I recall. I haven't used it since, but I think most of it has been cleared up.

Ps.
I found the thread where -h made that comment. Check here (http://forum.doom9.org/showthread.php?s=&threadid=22952).

[Teegedeck]

Originally posted by Acaila
Originally yes, but now they're of the opinion that using it for both passes would actually be better.

When it was first made unavailable in the 1st pass it still had quite a few problems as I recall. I haven't used it since, but I think most of it has been cleared up.

Ps.
I found the thread where -h made that comment. Check here (http://forum.doom9.org/showthread.php?s=&threadid=22952).

Checked it out. You're right, you CAN use lumi in the 1st pass, now. Upon closer examination, there haven't been any changes upon the way lumi-masking works in two-pass, except for that possibility. Basically, using lumi-masking in 1st pass _can_ (but needn't neccessarily) still result in a bigger base-quantizer for the affected frames when you do 2nd pass, which would be undesirable. The possibility though should be small, with the new (tamer) lumi-masking code. Try it out, I guess I will, too, sooner or later.

[jggimi]

I've just tried Light Psychovisuals with City of Angels. There are several scenes in it with the Pacific shore, including two scenes with U/W photography.

I didn't think anything of it, until I watched the whole movie....and noticed ... or didn't notice, any ugly blocks. None.

The water looked like .... water. And this on a single CD rip at 700kbps.

I'm thinking of testing Doom9's famous SPR beach scene he uses in his codec tests.

Has anyone else tried a water scene?

[Fox Mulder]

One thing I noticed is that using "light", "normal" or "strong" psychovisuals with DivX 5.0 Pro in 100% Quality Based, the results are the same size(byte identical) and with artifacting; and in 5.0.1 Pro the results are still identical in the "normal" or "strong" settings(smaller size than "light"), but without the artifacts.
And the "beta" testing continues...:rolleyes:

[-h]

Re: lumi masking in XviD.

Yes I will re-enable the checkbox, as lumi masking should most definitely be applied in both passes, if it's wanted.

I'll try to make the changes tonight. Hopefully it doesn't require a mod to the hint system.

-h

[Nazgul]

Originally posted by Fox Mulder
One thing I noticed is that using "light", "normal" or "strong" psychovisuals with DivX 5.0 Pro in 100% Quality Based, the results are the same size(byte identical) and with artifacting; and in 5.0.1 Pro the results are still identical in the "normal" or "strong" settings(smaller size than "light"), but without the artifacts.
And the "beta" testing continues...:rolleyes:

My guess is that you're not going to see a huge benefit from psychovisual enhancement in Quality-Based mode, especially 100%. Psychovisuals seem to be all about efficient distribution of bits, focusing them in the areas of the movie that they'll be noticed more. As such, it would make sense that it'll be more effective to use in situations where one is trying to maintaint a target bitrate rather than a target quality. I have a feeling that the folks who will benefit most from it will be those doing 2-pass or 1-pass(bitrate-based) at low to medium bitrates. Which would explain my interest in it, since my encodes tend to be 2-pass at rather modest bitrates. :)

[puffpio]

modest bitrates? heheh me too.
I'm always trying to cram two hour movies and up onto 1 cd.

My first test was an easy one: Batman, since it is a very dark source (although the DVD transfer is CRAPPY the video wobbles and you can see the film imperfections a lot)

I used a light psychovisual setting. Am gonna try 'normal' for the next movie

[Fox Mulder]

I see your points, but I found that DivX 5.0.1 Pro can fit a 2 hour movie in one 700MB CD lowering the resolution a tiny bit, and using 100% CQ, B-frames, GMC, and strong psychovisuals. I don't think my metod is better but I like the data reduction the MPEG4 tools provide, even using the min quatizers. I like this because all the frames get the same compression and it takes half the time.
And now you can flame me...:D

[unplugged]

Originally posted by Fox Mulder
...using 100% CQ...

100% CQ ???

What's it? :confused:

[dragoman]

but I found that DivX 5.0.1 Pro can fit a 2 hour movie in one 700MB CD lowering the resolution a tiny bit, and using 100% CQ, B-frames, GMC, and strong psychovisuals

Sometimes this is correct....but not always. Depends on your source movie and how compressible it is...

100% CQ ???

Refers to 100% Constand Quality mode (AKA Quality-based encoding) where you set a certain quantizer for the codec to use. 100% setting means that for every frame (except b-frames, the quantizer is doubled for b-frames) the encoder will use a quantizer of 2, achieving the max quality available.

dragoman