Hi,

The very firt results of this new version are very strange:
- MPEG2 Better than MPEG1
- MPEG1 Better than previous version
- CQ better than 2 pass VBR
- No more flashing artifacts because of I frames
- But some stranges artifacts: some lines disapears in moving frames

Test:

Max bitrate: 2100
Min bitrate: 64
Matriz: KVCD (couldn't be other!)
GOP Length: 25 (Pal world; sorry guys! : )
Motion estimation: High (better results than Highest and Standard)
Sample: 3968 frames
Resolution: 544x576

VBR: 1200
MPEG2: 23073 Kb/SSIM 62,14

CQ: 63
MPEG1: 21319 Kb/SSIM 64,66
MPEG2: 21815 Kb/SSIM 66,96

Artifacts: http://www.digitalfaq.com/archives/error.gif
I tried changing matrix: nothing
I tried changing Optimize half/full pixel motion search: nothing

Don't know, but as it's a moving scene, I have to see it in the TV to see if it's noticeable or not...

More news later.

Fabrice

@fabrice
Are those red lines from the encoded sample 8O ?
How did you manage to get TMPGEnc XPress to read the avs files? I've just downloaded Media Player 9 but can't see anything about avs files... Please, tell me what to do, I really want to test this new version of TMPGenc :wink: .

:lol:
It was to tell were to llok for the artifact !

You are too tired. Go to sleep a little :-)

@Dialhot
Yep, you're right again! :wink:
I'm very tired for the moment. Maybe I should come back here tomorrow when my head is clear again... :)

EDIT: What exactly are the artifacts? I don't think I see them (within the red square! :lol: )...

Hi,

To be able to read avs file, you have to activate readavs plugin in Option/Preferences/Input plugin settings. If it don't appear, just install readavs plugin.

The artifacts are that the lines are broken, and blurry. In the original vob, and in previousa version, there is no problem.

I'll desactivate every plugin except readavs, and encode sample, to see if it's better (o just leave directshow)...

EDIT: I got it: it was the 'desinterlace when necessary' checbox the responsible of this problem. (at list, at preview level)
More test this evening, as I have to redo everything...

Cu
Fabrice

Is there an easy way of importing the Notch Matrix or must I type it in by hand?

Hi audio slave,

Yes, there is an easy way; Just import this file in Set output tab, Custom Matrix editor/Open button.

Cu
Fabrice

Final results
 

Hi,

Well, I have at least encoded all kind of samples, and here are the values:

VBR: 1200 / CQ: 64
Resolution: 544x576

TMPGENC 3.0 (min 64 kbps / max 2100 kbps)
MPEG1:
- VBR: 23205 Kb / SSIM 63,01
- CQ: 23615 Kb / SSIM 65,84
MPEG2:
- VBR: 23138 Kb / SSIM 65,31 (with BV: min 148 kbps/ max 1978 kbps/average QLevel: 5,06!)
- CQ: 23458 Kb / SSIM 68,56 (with BV: min 223 kbps/ max 1815 kbps/average QLevel: 8,09!)

Tmpgenc 2.521 (min 64 kbps / max 2100 kbps)
MPEG1:
- CQ: 23646 kb / SSIM 61,53

Mencoder (using MencodeME):
MPEG2:
- VBR: 23324 kb / 66,51
Min with Bitrate Viewer: 158 kbps
Max with Bitrate Viewer: 2250 kbps
Average QLevel: 3,33

So you can see that with this movie (Beauty and Beast), Tmpgenc XPress is the winner... and that MPEG1 CQ mode is near mencode 2pass quality.

The script I used was a simple one: only resizing with Bicubic Precise. Nothing else.

Let's see samples...

C
U
Fabrice

It's just that the quantizers don't necessarily have anything to do with visual quality!

@fabrice
First, I would like to thank you for the Notch Matrix file :wink: .
Second, I agree with you about the quality of TMPGEnc XPress. It gives me encodings (MPEG-2) that looks much better than with every other encoder I've ever tried - even better than MEncoder! I did a sample clip of a scene from "Gothika" with T.XPress and the mosquito noise I saw in my MEncoder clip was no longer there! Amazing :D ! Now I seriously consider buying this piece of software... I'm really impressed with the quality this encoder produces.

Good job testing Fabrice,seems for promising

Hi Audioslave have you done mpeg1 yet ,and if so how were the results,also we still will have the predictioin problem .

Correct?

@bigggt
Hi, no I still haven't done MPEG-1. I've just bought myself a DVD buner so my first priority was MPEG-2 :wink: . But, I will make a test encode of the same clip in MPEG-1... In a couple of minutes as a matter of fact! :)

And, yes, we still have the prediction problem with TMPGEnc XPress (I think) :cry:

As fabrice said: MPEG-2 looks better than MPEG-1 8O ! And MPEG-2 looks unbelievable good :lol: .

Thanx Audioslave

I have been leaning towards a dvd burner but was trying to wait for new computer but with everything going Mpeg2 around here i'm not sure i can wait much longer

Mpeg2,a whole new learning experience :roll:

Thanx again

Has anyone tried CQMatic with TMPGEnc Xpress :?:
I wonder if they have fixed anything with the CQ linearity problems :roll:

-kwag

Hi,

To do the test, you have to buy it, because the trial version don't allow you to save projects....

I'll encode a sample at 3 different CQ, and tell you if it's a linear curve or not (67, 70 and 75, for example). If it's not linear, we still have the prediction problem...

CU
Fabrice

:x Thanks Fabrice :D

-kwag

Hi,

Bad news: still not linear...
CU
Fabrice

Thanks Fabrice :!:
That's what I thought. I guess they just tweaked their internal engine somehow, but it's still basically the same encoder, just more optimized for MPEG-2

-kwag

Well,

But as 2Pass vbr and mencoder 2pass have similar quality, you can use tmpgenc vbr, with a average bitrate, and still get the same quality than mencoder.

[ad mode]
What's the advantage? The image with Tmpgenc is more 'stable', and has got less flying artifacts.
[/ad mode] :wink: (Remember: I don't work for Pegasys! :P )

But remember: I'm still doing KSVCD, so with High bitrate, don't know how is the quality...

Any Volunteer?

Cu
Fabrice