Testing CQMatic Versions
 

I think there should be a thread where we all collect the results of Our testings.

I think this is the best way to post this:

My girlfriend cutted of the power yesterday.... :cry: :cry: :cry: :cry:
So I 've to start again...

Re: Testing CQMatic Versions
 

I assume that's for a 90 minute CD-R (800MB + Audio :!: ) :?:

-kwag

Hi Kwag,

sorry for posting so late. Been very busy last days :roll: .

I've just predicted a movie with ToK and CQMatic 0.92 b2a.
Movie details: PAL, 114 minutes, Average Bitrate 851
I'm doing a full encoding right now, results in about 10 hours.

Thanks Kwag for developing this tool :!:

@krassi a little bit less information about Your movie... 8O

... but it is ok!

Wanted video filesize by Moviestacker: 704 MB
Final size with CQ=49,38: 624 MB.

Hi Krassi,

Try version 1.0 and see if you get a higher CQ :!:
Please tell me what you get. Just run prediction only.

-kwag

Yep, i'll do this this evening.
The funny thing is that ToK-Prediction was even worse. Thats first time ToK was so way off... I'll try another movie tonight where i already know a right CQ...
Has the problem with short movies been solved?

How off was the CQ in ToK compared to CQMatic :?:
And the problem with short movies should now be solved with Version 1.0 ;)

-kwag

ToK predicted a CQ of 45,507 (see above post).
But i've done only a short sampler with ToK. This morning i started a new prediction with ToK with a bigger sample size. Maybe then its more accurate.

I'll try now with a shorter movie (i have one @work) and newest CQMatic. :D .

Short movie results (23 minutes):
Wanted size: 188.61 MB
Final size: 161 MB
I think thats not bad for such a short movie :!:
This evening i'll make a test with a longer movie.

@Kwag:
Can you explain me (shortly) how your prediction algorithm works :?: Or is it the one from the "File prediction experiment #1,000,000 :)" -Thread :?: Just curious :book:

hi kwag me again :)

testing new cqmatic version :

black knight (91 min)

predicted a cq of 66,81

well lets see whats coming out :)

Compared with ToKs result:

Hi Krassi,
Take a look at the "Wanted" size in moviestacker for both encodes, and compare that to the size created by CQMatic and ToK. (Not the wanted from ToK's log )

-kwag

Hi Kwag,
filesize in Moviestacker was 188.61 MB for both.

Ok, so it seems I'm under target. Not much, but it's still below. I just need to find out if it's the same behaviour for everyone. If not, then I have to go into another file prediction method.
However, black prince reported an extremely accurate encode with a movie, so if this is the worse case scenario ( minus ~15MB to minus ~17MB ), then I won't touch prediction :!:
That means that our encodes will be from ~exact size to a max of ~-17MB, and I would call that a very good prediction :)
I'd rather have that range, than some +20MB over the limit encodes.
I do have the prediction percent bracket in CQMatic set to -1% and +0.5% for a max range of 1.5% difference. This way, if my sample calculation is over the internal reference wanted sample, we ensure that the file size will not be that much larger than the wanted size.

-kwag

I think with that small sample size this is really good :!: I've configured ToK to take a bigger sample size...
I'm doing a full encode right now.

Im my case it was 35 MB under it!!!!

Version 1.0 beta3:
Same movie as before (23 mins).
Wanted filesize: 188.61 MB
Encoded filesize: 170 MB

We're getting closer :D .

Yesssss :D
That's VERY reasonable for the low side. Now I have to confirm that the "High" side is not too over the "Wanted" size. I'll encode K-Pax, whick was always over the size with ToK and the manual predictions with Sampler() or SelectRangeEvery() functions :)

-kwag