08-17-2003, 11:39 AM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by vhelp
Hi Kwag..
Incorrect
I chose ES (Video + Audio) for two reason:
A - because memory serve me that you (after testing) found that v2.520
......was ok w/ this mode and ES (unless I miss-read you )
B - because I want to save time, and not have to encode audio separately.
......if its already there, why not.. right ?
Ok, let me have it
Tell me, that I have to re-do again, w/ same CQ's ??
-vhelp
|
The average bitrate and movie time formula in CQMatic are designed for Video Stream only
I haven't made any test encoding audio with TMPEG, because the quality of TMPEG's audio encoder sucks
That's why CQMatic was developed for video, and not for audio
-kwag
|
Someday, 12:01 PM
|
|
Site Staff / Ad Manager
|
|
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
|
|
|
08-17-2003, 01:13 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Hi Kwag.. others..
FWIW...
I'm now runing my first predition test w/ my 2nd pc. Finally got most things
sorted out, and I'm now able to read (again) .d2v source files into TMPG.
I'm using ES (video only) this time around for this project.
I'm currently doing "Red Planet", but as soon as that is finished, I'll do my
"Blue Streak" again.
I'll let you all know how it fairs, ..hopefully better than my MAIN pc.
-vhelp
|
08-17-2003, 01:38 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
..update
* Movie: "Red Planet"
* CQ of 56.52
* MPEG-2 / 16x9 / 704x480
Will try "Blue Streak" now. I found out that I had Scene-Detect on, in this one.
-vhelp
|
08-17-2003, 01:48 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by vhelp
I'm currently doing "Red Planet",
|
Telepathy
I just finished "Red Planet" about an hour ago
Encoded at 352x480 directly from .d2v (for testing CQMatic 1.1.12a at that resolution).
Wanted size: 726,338KB
Final Encoded size: 740,083KB
For +1.8% diff.
Encoded at CQ=79.51 (By CQMatic)
Wana see what it looks like, with "Zero" filters at all , just KVCD's Q. Matrix
http://www.kvcd.net/rp-352x480-cq-79...0-max-2500.mpg ~10 second clip (no audio)
-kwag
|
08-17-2003, 01:55 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
.
.
Zero filters is the only way to go (if you can afford it space wise.. that's
why CQM's for )
D/L'ing sample now.
Maybe I should re-do "Red Planet" again, w/ 352x480 this time out of
curiosity. I expect to see some differences though, cause after doing both
test predicts on each machine, I found color space was off by one tick.. and
that means different CQ values (for me anyways)
Right after I complete predict on "Blue Streak" I'll give "Red Planet" a go !!
-vhelp
|
08-17-2003, 02:00 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Sample looked very good
Glad to see you have (seemed to have) finally met up w/ the boarders
..no blocky-bleeds IYKWIM !!
Hay, it would be interesting to see what my MPEG-2 16x9 704x480 would
look like, after CQM gets to it w/ a CQM ..for comparison sakes.
Can't wait to try that out
-vhelp
|
08-17-2003, 02:02 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by vhelp
Hay, it would be interesting to see what my MPEG-2 16x9 704x480 would
look like, after CQM gets to it w/ a CQM
|
It will look good, but just not quite as good as the MPEG-1 version
-kwag
|
08-17-2003, 02:20 PM
|
Free Member
|
|
Join Date: Jun 2003
Location: Orange County, CA
Posts: 291
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by kwag
Quote:
Originally Posted by vhelp
Hay, it would be interesting to see what my MPEG-2 16x9 704x480 would
look like, after CQM gets to it w/ a CQM
|
It will look good, but just not quite as good as the MPEG-1 version
-kwag
|
I have found you don't need to use the 16x9 setting in TMPGEnc, but just have to set dest_anamorphic=True in gripcrop. If your source is anamorphic this will give you 16x9 output which looks great on a widescreen TV.
Downside is that on 4:3 TV it won't look quiet right and people will be squished and look skinny and stretched. Also since you are using more resolution you'll get lower CQ values.
On that note I use dest_anamorphic=true for my encodes
|
08-17-2003, 02:58 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Hi audi2honda,
That might be true, but with the method I use now, and for 16x9 output,
weather I view on a 4:3 TV or 16x9 TV, the output will always look the
same !!
MPEG-1 can benefit this as well, and save on some bitrate/CQ and things,
if they choose 16x9 (but is tricky in some situation ie, kinds of TV and SW
viewing) But, you can cut your filtering by a lot if you use 16x9 as your
output, OR, resize properly, and totally eliminate the boarders in your
encodes, if they are with in measurement and AR calculations.
Also, if you dvd2avi w/ RGB and select TV as your final .d2v process,
you'll get better color space quality in your final MPEG-1/2 encoding.
I've outlined this on this thread somewhere in the begining. You should be
able to copy the steps as is. I have found this to be the Ultimate in final
quality. It's the best, and is what I use for maximum color quality. See
for yourlself, when you play it on your TV. Don't alwas go by what your
eyes see on your PC monitor. I've learned this well over a year ago, burn
to CD and test your final analisys. You'll see (provided you processed your
project just right) There's a balance w/ everything. The wrong key, can
throw off your final results. My LCD monitor can pick up quite a bit of
detail. I can usually tell how an encode was done, just by looking at the
samples on my LCD screen. Its a great tool (I've said this elsewhere)
In any case, these are my recommendations for best quality output.
See Kwag's latest RedPlanet sample for an example of how far he
has gone this technique (give or take a little) you know how much
he loves MPEG-1 no matter what hehe
A tip from me is, if you can afford it, get rid of the boarders altogether.
They are a waist of bitrate period. You only need to figure (learn) how to
obtain the corred (or near) AR values and resizing technques. Something
that I'm STILL figuring out myself
I'll try and show some sample clips of my MPEG-2 16x9 encodes of my
work, that I've ben using for a while now, ...when I get the chance to U/L
them. These play great on 4:3 TV (standard sets) and resize the 16x9 AR
for proper viewing and WITH boarders, ...but at the moment, I'm waiting
on testing true 16x9 widescreen TV sets - - I need one, as no one here
seems to want to let me know how mine come out on theirs.
My other issues are related to my process. Rather what settings I use. ie,
I had scene detect enabled in past predicts and encodes. These can throw
off very easily final results.
At the moment, I'm having one of those issues, unfortunately
-vhelp
|
08-17-2003, 03:14 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@ Kwag..
* Movie: "Blue Streak"
* Length: 94 minutes
* Audio: 112k
* Max: 2000
* Ave: 1050
* Min: 599 ... (by vcalc)
* MPEG-2 / AR: 16x9 / Res: 704x480
* Scene Detect: unchecked
* GOP: 1,5823,2,1,24
* [x] Closed GOP
What's with this darn crawling hehe. It's driving me madly insain
I just wanna end the whole process altogether when it does this.
Using CQ of 60.00
Current CQ = 47.24
Current CQ = 43.76
Current CQ = 42.74
Current CQ = 41.78
Current CQ = 40.70
Current CQ = 39.75
Current CQ = 40.22
Current CQ = 39.98
Current CQ = 40.10
and, an hour later...
If you could add a text box [.....] so that users who suffer this crawl
can input a "bump" value, that might help reduce the:
* number of passes, and
* length of time to predict
A little intervention woudn't hurt here, I don't think.
As a user, and seeing the CUE ie, cq 47.26 and cq 43.76, that
would be my CUE to bump the CQ value, say by 5 ie, 47.26 to 42.26 or
something like that. Than see where CQM goes from their. But, instead
of waiting 1/2 hour or more for, and 5 ore more passes, and I would have
finally arrived at my final CQ anyways !
Otherwise, this would drive me waiting, when I already know what
the final CQ value will be (usually very low)
At these low values (if lower than usual) I would normally know what to expect
in quality w/ these numbers. In my expeirence, CQ 40 would be like,
the lowest I would go in CQ.
Does this make any sense so far ??
-vhelp
|
08-17-2003, 04:24 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@vhelp,
There's no way around it
The problem is the almost "Flat" line around CQ of ~40
Where a large change in CQ causes a very little change in file size. So I need many passes to find the optimal CQ. If I add an override option, then I'll get feedbacks like "My file size was way over target" or "My file size was way undet target", because of users overriding the CQ value. So I'd rather wait an hour, for 5 or 6 hours of encode, and finish with a very close file size, than Guesstimate a CQ, and wind up with a file size way off. Remember, a little file size difference on the sample, can mean a HUGE file size difference on the full encode. That's why it's so critical on the small size sample being scaled
-kwag
|
08-17-2003, 05:26 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Hi Kwag..
Sorry for not responding sooner. Lightning struct one of our facilities, and
notcked out everything. No power for a while. Lost my last CQM testing
and was alost most there.. 8th pass. Shucks
Quote:
So I'd rather wait an hour, for 5 or 6 hours of encode, and finish with a very close file size, than Guesstimate a CQ, and wind up with a file size way off. Remember, a little file size difference on the sample, can mean a HUGE file size difference on the full encode. That's why it's so critical on the small size sample being scaled
|
Ok I understand, and my head does too
I did a small sample clip of "Red Planet" earlier, ..same scene as your sample,
but using my process and settings and things.
Anyways, I'll U/L them, case you wanted to see what I've encoded, based
off what I discussed previously here. As always, your input is more than
welcomed
I'll give "Red Planet" specs in a moment, followed by a sample clip based
off of CQM's result.
-vhelp
|
08-17-2003, 05:45 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@ all..
For those who were following, as promised, below are my reading for
CQM v1.1.12a and a small sample clip based off those results.
Have a nice and remaining day all
-vhelp
Sample can be found here:
* SAMPLE Encodes..
Remember.. the sample was intended for PowerDVD users only.
Please read the PDVD notes on my SAMPLES.. thread
Below were my CQMatic v1.1.12a results:
http://www.kvcd.net
CQMatic Version 1.1.12a
Copyright Softronex Corporation, 2003.
All rights reserved.
Time: 18:02:20 Date: 08/17/2003
Ready!
Project: C:\Program Files\Pegasys Inc\TMPGEnc Plus 2.5\v2.520-redplanet.107.cq(50x2000x51 .av909.cqm(v1.1.12a). tpr
Creating: CQMatic.tpr
e:\2nd.redplanet.rgb.tv.1.4.m2v
Project resolution: 352x480
Execute.
Movie Time: 107
Average Bitrate: 909
Prediction Only mode
Executing Prediction Phase...
Process started at 18:02:49
On 08/17/2003
CQ set for prediction
Setting up initial sampling.
Using CQ of 60.00
Prediction cycle #1
Encoder started...
Process time: 3.92 minutes.
Encoder end.
File size difference = 1.312513
Low fence: 60.000000
High fence: 90.000000
Last CQ = 60.00
Current CQ = 78.75
CQ difference = 18.750755
Using CQ of 78.75
Prediction cycle #2
Encoder started...
Process time: 3.98 minutes.
Encoder end.
File size difference = 0.907897
Low fence: 60.000000
High fence: 78.750755
Last CQ = 78.75
Current CQ = 69.38
CQ difference = 9.375374
Using CQ of 69.38
Prediction cycle #3
Encoder started...
Process time: 3.92 minutes.
Encoder end.
File size difference = 1.148463
Low fence: 69.375381
High fence: 78.750755
Last CQ = 69.38
Current CQ = 74.06
CQ difference = 4.687683
Using CQ of 74.06
Prediction cycle #4
Encoder started...
Process time: 3.93 minutes.
Encoder end.
File size difference = 1.035555
Low fence: 74.063065
High fence: 78.750755
Last CQ = 74.06
Current CQ = 76.70
CQ difference = 2.633278
Using CQ of 76.70
Prediction cycle #5
Encoder started...
Process time: 3.97 minutes.
Encoder end.
File size difference = 0.929733
Low fence: 74.063065
High fence: 76.696342
Last CQ = 76.70
Current CQ = 75.38
CQ difference = 1.316643
Using CQ of 75.38
Prediction cycle #6
Encoder started...
Process time: 3.97 minutes.
Encoder end.
File size difference = 0.969977
Low fence: 74.063065
High fence: 75.379700
Last CQ = 75.38
Current CQ = 74.72
CQ difference = 0.658318
Using CQ of 74.72
Prediction cycle #7
Encoder started...
Process time: 3.97 minutes.
Encoder end.
CQMatic complete!
Total minutes of process: 27.72
Process ended at 18:30:32
On 08/17/2003
|
08-17-2003, 06:08 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Looks great vhelp
But your aspect is not correct
You encoded as "Anamorphic" output, instead on non-Anamorphic.
That would be ok if your target was DVD, but it's not the case, after seeing your VBV buffer size
-kwag
|
08-17-2003, 06:13 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Thanks Kwag, for your comments
Yes, I realized that. And, I probably should have changed it. But, I was
testing, so it didn't matter so much.
Now, I'm going to try a 704x480 (DVD intended) using same (modified)
settings as prevous one used for the 352x480 above. I'm curious to see
what CQ I get for this one.
I'm contemplating on encoding the 352x480, just to see if it all goes on one
800mb CDR though.
Well, let me get to another "Red Planet" predict.
-vhelp
|
08-17-2003, 09:10 PM
|
Free Member
|
|
Join Date: Mar 2003
Location: IamCanadian
Posts: 848
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
divx movie
Quote:
http://www.kvcd.net
CQMatic Version 1.1.12a
Copyright Softronex Corporation, 2003.
All rights reserved.
Time: 13:09:18 Date: 08/17/2003
Ready!
Project: C:\Temp Videos\old.tpr
Creating: CQMatic.tpr
C:\Temp Videos\old.m1v
Project resolution: 528x480
Execute.
Movie Time: 92
Average Bitrate: 1083
Full Encode mode
Executing Prediction Phase...
Process started at 13:15:00
On 08/17/2003
CQ set for prediction
Setting up initial sampling.
Using CQ of 60.00
Prediction cycle #1
Encoder started...
Process time: 9.97 minutes.
Encoder end.
File size difference = 1.468953
Low fence: 60.000000
High fence: 90.000000
Last CQ = 60.00
Current CQ = 88.14
CQ difference = 28.137169
Using CQ of 88.14
Prediction cycle #2
Encoder started...
Process time: 10.40 minutes.
Encoder end.
File size difference = 0.662710
Low fence: 60.000000
High fence: 88.137169
Last CQ = 88.14
Current CQ = 74.07
CQ difference = 14.068581
Using CQ of 74.07
Prediction cycle #3
Encoder started...
Process time: 10.27 minutes.
Encoder end.
File size difference = 0.932752
Low fence: 60.000000
High fence: 74.068588
Last CQ = 74.07
Current CQ = 69.09
CQ difference = 4.980988
Using CQ of 69.09
Prediction cycle #4
Encoder started...
Process time: 9.58 minutes.
Encoder end.
File size difference = 1.210207
Low fence: 69.087601
High fence: 74.068588
Last CQ = 69.09
Current CQ = 71.58
CQ difference = 2.490494
Using CQ of 71.58
Prediction cycle #5
Encoder started...
Process time: 9.55 minutes.
Encoder end.
File size difference = 1.021302
Low fence: 71.578094
High fence: 74.068588
Last CQ = 71.58
Current CQ = 73.10
CQ difference = 1.524788
Using CQ of 73.10
Prediction cycle #6
Encoder started...
Process time: 9.18 minutes.
Encoder end.
File size difference = 0.958535
Low fence: 71.578094
High fence: 73.102882
Last CQ = 73.10
Current CQ = 72.34
CQ difference = 0.762398
Using CQ of 72.34
Prediction cycle #7
Encoder started...
Process time: 9.38 minutes.
Encoder end.
File size difference = 0.984019
Low fence: 71.578094
High fence: 72.340485
Last CQ = 72.34
Current CQ = 71.96
CQ difference = 0.381195
Using CQ of 71.96
Prediction cycle #8
Encoder started...
Process time: 9.35 minutes.
Encoder end.
File size difference = 0.995726
Low fence: 71.578094
High fence: 71.959290
Last CQ = 71.96
Current CQ = 71.65
CQ difference = 0.307564
Using CQ of 71.65
Prediction cycle #9
Encoder started...
Process time: 9.65 minutes.
Encoder end.
File size difference = 1.015735
Low fence: 71.651726
High fence: 71.959290
Last CQ = 71.65
Current CQ = 71.81
CQ difference = 0.153786
Encoding set to Full encode.
Full encode start...
CQMatic complete!
Total minutes of process: 87.33
Process ended at 14:42:20
On 08/17/2003
|
wanted vs-738220
encoded vs-799565
7.6%
|
08-17-2003, 09:12 PM
|
Free Member
|
|
Join Date: Jun 2003
Location: Orange County, CA
Posts: 291
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Yea 1.12 is going way over for me too on my 70%film/30%video hybrid material. I don't have any exact numbers but my last 3 encodes have been 35-50MB to big for 1CD. I've gone back to 1.11 for now.
Haven't tried any pure progressive 100% film sources yet.
|
08-17-2003, 09:25 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
.
.
Well, I'm doing "Red Planet" as we speak.
With 4 hours to go.
* MPEG-2 / 16x9 / 704x480
* CQ: 50.88
* 1 CDR @ 800mb
Here's what I have, based on my new modified HD size reader app
for Windows 98, for the currently encoding project file:
Size: - - - - - - Timed - - - - - Position in Movie:
------------------------------------------------------
172mb - - - - - 1h:07m - - - - 25 minutes
Can anyone do the math on this one ??
Will it make the 800mb 1 CDR goal ??
Quality looks pretty good, even at CQ 50+
-vhelp
|
08-17-2003, 09:27 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Ok, I really need to know if accuracy of 1.1.11 or 1.1.12a is better. However, all reports must be consistent. That is, if everyone gets great results with 1.1.11, then that sampling is the one to be used on future versions. 1.1.12a is different. It has a longer sample window, but takes far less samples per movie. I'm still testing both, with mixed results
-kwag
|
08-17-2003, 09:39 PM
|
Free Member
|
|
Join Date: Jan 2003
Posts: 1,009
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Hi Kwag..
I'm not sure what you mean by consistant, but..
Count me in, for my "Red Planet" project, if you can wait till 4 hours later
for the results.
You already know the stats for that movie But, see above for my
setup for CQM v1.1.12a
I don't know if all these are factors in your analisis, but I'll help if I can.
Boy, am I glad I got my 2nd pc up and running again for these kinds of fun
things, while I do other fun stuff on my main pc
"Red Planet" encode project:
Size: - - - - - - Timed - - - - - Position in Movie:
------------------------------------------------------
216mb - - - - - 1h:27m - - - - 30 minutes
Does it matter for:
* source type ie, divX or DVD rip or Capture
* NTSC vs. PAL
* Film vs. Interlace
* Movie length
-vhelp
|
All times are GMT -5. The time now is 01:24 AM — vBulletin © Jelsoft Enterprises Ltd
|