trying now with the newest version of TMPG, but if i see that the cq is the same , i cancel.
with rc2 , i have the feeling that i had to set the cq value manually down . f.ex, prediction is 74, i´d use then 72, or 71 sth like that, then it would be (perhaps) fitting all the way :) lets see, going on testing ;) |
I've been using the latest tmpgenc but come to think of it before I was using 2.513 when I was getting accurate cqmatic predictions. Just noticed that kwag put a note that cqmatic is not accurate for avi files... I'm taking the same approach as you for testing, if I get the same CQ I know it's not accurate. Right now I'm going back to beta 4 and from the start of things, it's going down in CQ! It'd be interesting to prove whether some of the earlier betas were actually more accurate for avis or not.
|
i´m not using avis, i have dvd as sources well....
now im testing rc3 again, with NEWEST tmpg version, and other settings , lets see later... |
cancelled test, cq pretty the same...
another test now, taking a cq of 77,31 , with that the video stream was 727mb , so a little big, will take now 74,31, lets see results later bye |
Quote:
-kwag |
i just don´t know further :(
i can use EVERY setting for max and min bitrate, butt the vs is always gettin 20mb-30mb too big :(( could any1 up the beta6? with that i had the most success |
JUst my opinion about the newest tmpge 2.520 ,can't say for sure (not enough tests) but so far even with tok the video seems to be too much.When using tmpge 2.58 if anything the video would be too small.I did a test with the newest cqmatic and got a cq of 81 and got a video size of about 798mb just video,and with tok the cq was almost identical but got a video size of about 776 just video.(the sources were avi's).I am going to run the same avs through tok with tmpge 2.58 and see what happens
|
Hi tigger,
These is my setup, just in case there are some differences to what you are using: TMPGEnc Plus version 2.520.54.163 AviSynth 2.52 Build July 17, 2003 [12:46:12] All my tests we using the MA script, as is, and also without any scripts at all. Just reading the .d2v directly with TMPEG. Here is a project file, which you may copy and load in TMPEG. Then change the relevant settings for your project, and save the project again: Code:
object TMPEGEncodeJobFile So far, AVIs are predicted incorrectly. I posted a note at the top of this thread, related to that. All my tests have been with DVD sources. Apparently, it has to do with the way prediction samples the source, so I believe it's a CODEC issue. Right now, I'm strictly working on .d2v (mpeg) sources, so I can tune precision to the max. Then I'll proceed to figure out why the strange behaviour with AVI sources. -kwag |
Does anybody have any answers to my previous post?
Quote:
Will there be a user guide out soon? |
Quote:
I didn't see your post 8O You don't need avisynth. As a matter of fact, you can even use TMPEG directly on your source, but you won't gain the advantage of using avisynth's filters :) You can use any bitrate calculator. Quote:
Quote:
Look on "Avalon's" signature. -kwag |
Quote:
I didn't see your post 8O You don't need avisynth. As a matter of fact, you can even use TMPEG directly on your source, but you won't gain the advantage of using avisynth's filters :) You can use any bitrate calculator. Quote:
Quote:
Look on "Avalon's" signature, or here: http://www.kvcd.net/forum/viewtopic.php?t=5188 -kwag |
RC-4 is up :!:
As usual, link and change log here: http://www.kvcd.net/forum/viewtopic.php?t=5145 -kwag |
@Kwag,
As promised, here's my results of "Green Mile": Movie .................... Green Mile Length in Time ....... 191 min 55 sec ( 3 hours 11 min 55 sec) Template ............... 352x240 ULBR KVCD Average bitrate ...... 508 (MovieStacker) Min ....................... (508 x 0.57) ~ 290 (Tmpgenc) Max ...................... 1150 (Tmpgenc) Audio size ............. 64 kbps (~90mb) (MovieStacker) Video target size .... 709 mb (MovieStacker) CD ....................... 1 80minute CQMatic ................ version 1.0 RC-3 CQ ....................... 79 (CQMatic) CQMatic log: Quote:
Final Video filesize ........... 694mb Final Audio filesize ........... 90mb Total A/V ........................ 784mb Comments: This was really a test for filesize prediction accuracy which is only 16mb under the prediction. That's very accurate :D The video took ~9 hours to encode with credits included. It's looks great on a small screen and only during action scenes is there light blocks and Gibbs effect is very reduced. The only drawback was the lengthly time (2 hr 6 minutes) it took for CQMatic to get the final CQ. I feel confident that CQMatic will give me a good filesize prediction if the source is DVD. :wink: -bp |
Quote:
@kwag Does CQMatic make it's prediction on the whole movie or just a short sample? Would SVCD still give a better result over VCD when using CQ encoding? Cheers |
Hi Kwag,
CQmatic seems so easy to use :D I did have a question in why is the min/max values so important in determining final CQ and file size accuracy? It could of been posted previously - though fell behind in reading the posts the last couple of days (couldn't get on KVCD.net -- what up with that Comcast :? ) Anyway, using RC-1 Kanagaroo Jack 352x480 Cq=66 min=400 max=2500 file encode size was right on :P Now I tried encoding the same movie at 480x480 (can't encode at full KVCD-3 resolution .... DVD player limited) CQ=57.5 min=.57*average(=630) max= 2000 file size +13M (too big) CQ=57.5 min=300 max= 2500 file size -12M (too small) CQ=57.5 min=400 max= 2500 file size -7M (too small - actually perfect once I burn it on 1 CDR) It's almost like this min/max is the fine adjustment for a specific resolution. Why the variance in file size by changing the min/max setting? Now to read more posts Thanks |
Quote:
Quote:
Quote:
For 352x resolutions, CQ_VBR provides better results. In both cases, it applies to both MPEG-1 and MPEG-2 -kwag |
Quote:
The problem is that CQ between ~75 to 90 is not linear. That's why it took so long. Same applies for CQ between 1 to ~50. So the best range will always be wneg CQ is between ~50 to ~75 If TMPEG's CQ was linear, it would only take one prediction encode :D I'll see what happens when I add support for CCE and MCE :cool: -kwag |
Quote:
It has to do with the dynamic range in TMPEG. If average bitrate is exactly centered between MIN and MAX, that's excelent for the encoder. In our case, we usually are encoding with a very low average, closer to MIN than to MAX. This throws off the dynamic range of the encoder :) Quote:
Quote:
Quote:
Quote:
-kwag |
@Kwag,
Kwag wrote: Quote:
Can't wait for the CCE option. Finding the Q value determines filesize. The lower the Q, the higher the quality and filesize. For MPEG-2, CCE has the speed and is quality wise the same or better than Tmpgenc :) -bp |
Quote:
You can even just do this: Set CQ=0 for long movies, and set it to 100 for short movies. CQMatic will adjust the starting CQ (or CQ_VBR) for you automatically ;) -kwag |
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.