@tigger,

Ok, I just did a test on a 57 minute segment, and CQMatic worked flawlessly :)
Starting CQ=50. Then jumped to 90, and then full encode at 90. That's the way it's supposed to work, so I assume that's the way it also worked for you. Check your log, and see if it reads something like: CQ ABOVE watermark. CQ set to 90.0
So it actually did a first prediction at 60, then a second prediction at 90, and then started the full encode at 90 :)


-kwag

yes , the log says this : CQ ABOVE WATERMARK SET TO 90...

well, i´ll see the final file size, hope its not TOO big now, we´ll see :)

after this, will do another film with cqmatic, will post every result to fix bugs and so :)

bye :)

What I'm looking for is consistent results :!:
If say, everyone gets the same results, either below or above wanted file size, then I'm happy as I can correct the prediction factors. Right now, as it is, I'm getting final size just under the wanted size. I'm currently encoding a movie ( which was ~30MB low ) without any filters. If I get the same (or almost the same) encoded size, then I'll be even happier :)
I believe the way I'm doing prediction is far more precise than the other methods being used. To give you an example, I'm sampling the material every 15 seconds, for .25 seconds. That is, about 6 frames per sample. So I'm sweeping the movie with a higher granularity, instead of one second every minute, as with the usual prediction method. Not to mention that I'm not using AviSynth's functions for prediction at all :!: It's all done directly in TMPEG :)

-kwag

well encoded finally my 53 min movie, it has a size of 843 mb :)

but i already found my fault, in moviestacker, i used the wrong length value, well when i use the right one i think its gonna be ok :))

bye

What did you have set :?:

53 min, but its 53 min 31 sec

hmm...

cqmatic is now working and encoding smaller movies, but the filesize of the videostream is always getting too big...
i checked moviestacker but everything is ok...

cqmatic is always takin 90 as cq value , well but this value is too big, because the video stream is going to be 780mb-800mb, when i mux it with audi stream, its too big :)

what could that be?
perhaps i have to make a change?

I see that CQMatic adds B picture spoilage = 20 in the Rate Control Mode, even though it is set to 0 in the project. Should it do that?

No it should't. Jorel reported that to me yesterday, and today I'll take a look. Not that it makes a notable difference in quality!, but it should be set to 0.

-kwag

I'll take a look at this right now. I must have a wrong logic flaw in the program for some high CQ conditions :!:
I'll encode a 57 minute clip I have, and see if I can reproduce the problem.

-kwag

i did the prediction now with using
CQMatic1.0b1 and after search for 8 times give
CQ66,32.

yesterday the 098 version give CQ66,64.

:!:
:wink:

edit:
i forgot..the source tested have 32 minutes.

I'm getting fine results with my 57 minute clip (average bitrate 1814) :!:
It did find that it had to encode at CQ=90, and it's currently at ~36% encode, and file size is 248,000KB.

I also finished two full encodes of a movie, one with the optimal script, and another one without any filters at all (just resizing in the .avs) The file difference was :arrow: ~3 MB :!:
This pretty much confirms that prediction is accurate, in the sense that it is producing a constant file size. Wanted final size is 767,644KB, but it might not be reached, which is normal because of the MAX setting of 90. But it should never go above the wanted size :!:

-kwag

Shit... this new version doesn't work... where can i find the older version to check, if it's a problem with this new version...

My Problem: The CQ is always 90 or CQMatic trys to find the best CQ by testing always the same CQ... that means it's an endless loop....

I've just found another problem... Now, CQMatic doesn't try to find the right the CQ... it starts with the encoding the film with any CQ! Also I select Prediction Only

kwag, your´re just encoding the video stream right? later, you mux it with the audio...

encoded black knight, cq predicted war 66,81.
file size is 722 mb (audio + video)
yesterday i had 733 mb...:(

Hey kwag!
This BETA3 Version is much better!!!! I'd worked with BETA1... Now I haven't any problems!!!
But If there are going to be some, i'll post them!!!

Thx a lot!!!

Thanks fn86 :)
BETA 3 shouldn't get stuck on a loop anymore.

-kwag

If BETA 3 did indeed solve all looping problems, then the next BETA(s) will be prediction accuracy tweaks, until we're all happy :)
Then the "BETA" name will be gone, and 1.0 Release will be posted :cool:

-kwag

Did 1.0 beta 3 increase your CQ-values :?:
Looking forward to 1.0 8) .

Yes it will :!:
Final tweaks to find the optimal values for prediction :cool:

-kwag

Oh no...

There's still a problem...CQMatic doesn't try to find the right the CQ... it starts with encoding the whole film with the (first) CQ from the tpr-file, you understand?!

Help......

Mh...I have this problem with avi-files.... not with DVD-Rips....