I notice in Tmpgenc the Constant Quality (CQ) feature is set at 60 in the KVCD template I am using, how [if anything] would it negatively effect the final encoded file if I crank it up to 100?
Does it make the file bigger? or does it really balls something up? lol
Cheers
+ Reply to Thread
Results 1 to 9 of 9
-
-
It will make the file much bigger.
I think,therefore i am a hamster. -
Originally Posted by johns0
Is this feature dependent on the source file?, by that I mean like most all things in this game is it the case that [by and large] it WILL increase the filesize but NOT by the same amount/percent/whatever in every case?
Has there ever been a poll carried out here to see how many members have thrown their PC at the wall or came close whilst trying to figure this lark out?
LOL -
[quote="junkmalle"]
Originally Posted by boing
I think I am going to adopt the approach that 'as long as I get a working disc with my files on it then quite frankly I am happy', or simply put I am running out of bloody aspirins and scotch! -
Constant Quality mode is for when you want to assure the quality of the result and don't care about the file size. Of course "Quality" in this sense means how closely the output matches the input. If your input is crap it won't get any better!
I think you'll find that the numbers start going up a lot more when you go over 80 percent.
I ran a little test clip of a ~10 second still scene that came out to about 3.2 MB at Q60. At Q80 it was 3.5 MB. At Q90 it was 5 MB, at Q99 it was 9 MB.
A similar length high motion sequence was 3.2 MB at Q60. At Q80 it was 5.6 MB, at Q90 7.8 MB and Q99 10.9 MB.
Note that each of those numbers includes about 0.5 MB audio. -
Originally Posted by junkmalle
Still 80 doesnt seem to be to much larger and the quality is miles better -
yayz! I remembered my password (actually just passing thru to see if my old SM2500 supports +R, or -R, before I make the big leap and finally get a writer - if it doesn't work with -R DL, then it's time for a new player too
)
EDIT
yes ... its making my head hurt to look at as well. Not least because of the narrow columns...... how to stretch?!
Oh and the rambling, but if you think i'm editing that now, you've got another thing coming
I think the following is fairly accurate, but having written it and all, i'm no longer certain it actually answers the damn question! Great eh!
EDIT
How CQ works is that it tries to compress the picture such that it keeps a continuous average "quantizer" (whether it's full scene or in every last block i'm not sure, but it has the ultimate effect on the fullscene quant anyway), roughly equivalent to the perceived image quality and (non-linearly) determined by the set CQ value, across the whole of the film --- where it's possible to do so within the constraints of the max / min bitrate of course.
(a higher quantizer, from 1.0 through up to 31.0, looking steadily more "compressed" or distorted in that nasty blocky / squelchy fashion)
i.e. if the CQ value asks more of the encoder than can be satisfied by the maximum bitrate, then it sort of kicks back into CBR mode and gives you whatever quality is possible with that bitrate, which theoretically, in unrealistically extreme cases, could be all the way up to the worst-looking 31.0 (basically, the picture will be built out of totally flat 8x8 squares) regardless of your CQ.
On the other hand, if the material (and the encoder) can easily satisfy even the minimum bitrate at your chosen CQ, then a similar thing will happen, but with the quantiser dropping right down to 1.0 (effectively "perfect" compression, within the boundaries of what's possible with MPEG)... and then holds it there if even that is "too much", making up the rest with padding. This is actually far more likely than the level 31 effect! And, I suspect, may be the basis behind what you are seeing.
Quite possibly, thanks partially to the characteristics of whatever it is you may be encoding (say, making a heavily letterboxed film - large swathes of the screen being virtually data-less black bars), the efficiency of the TMPGEnc encoder, and the perculiarities of the KVCD notch matrix, you are actually hitting 1.0 quantiser for a lot of the film, even though at CQ60 the number should be higher, probably around 4 or 5. (Once upon a time, when I was actively encoding a lot of VCDs, I actually took an afternoon to do some quick clips and map the common CQ ranges out vs quantiser... can't remember any more, though! What I do remember is that there are some points where it changing the CQ by as much as 3 or 4 points did absolutely nothing to ANYTHING, the effect seeming to be more of a compound thing when you altered e.g. the P/B picture relative qualities and numbers of them).
Changing the level to 80 would alter the "aim" of the quantiser setting, where it would be trying to hit with moderately busy input video, maybe to around Q3.0.... however if you're pegged up against 1.0 on the minimum bitrate for a lot of the film already, I think you can already see how this won't change things by a great deal!
The more complicated parts of the film will look a bit better than they did before (though in most scenes, any Q below about 6.0 is very hard to tell from the original, on-the-fly, even with the high pixel magnification VCD offers you), and receive more data, but the "quieter" (in data terms) bulk will not alter at all.
Were this but 2 years ago I'd probably whip out some comparison graphs (made using the incredibly enlightening BitrateViewer) to show you what's going on, but .... well, it's not. Just imagine the data chart of your video over time with two lines - one for the theoretical perceptual quality, where lower = better (or perhaps, it would be better to term it perceptual distortion? higher=more distorted), represented by Q, and the other showing the actual bitrate on a per-second basis (call it.. i dunno.. K).
For........ what would be a realistic real life example for me... hmmm.... yes, taking an example of the full theatrical anamorphic widescreen version of The Lord of the Rings (film 1), rezzed down brutally to fit a 4:3 PAL screen (could have been worse - at least it's still getting about 160 lines, with only some very minor stretching). Thanks to the enormous black borders, and the surprising amount of standing around and "quiet" scenes between all the high-demand running around and fighting, all 160-plus minutes of it fit on one disc.... The graph showing Q being a mirror flat line on 1.0 for a lot of the feature, with K hovering near to or at minimum bitrate in those areas, and then spikes upwards in areas of action, like mountains. Well - foothills. The bitrate chart goes crazy, up and down, but the Q forms a straighter line a bit further up the chart, about 3.3... then in the most crazy areas, the K line plateaus out at maximum and the Q rises up again - but not as far as you may think.
(after all - it was little more than half-height video, and the max bitrate was a hair below SVCD, as that's what my DVD player allows for xVCD - meaning the very worst compression artifacts were very, very minor, even with an average rate somewhere in the 500k's)
For this one also, when the settings were right, changing the CQ by quite large amounts didn't have so much effect on the bitrate...... which was highly annoying, when I was trying to bring it down to squish onto the disc without reducing the audio further... (160k mp2 stereo is painful enough as it is!)! ...... and encoding it on a sub-gigahertz machine!
If you can get hold of Bitrate Viewer (there's a free version, that's a bit crippled but shows you the essentials - same as mine - if it's no longer on the web i could try to find it for you) then I'd heartily reccommend opening the two files in it one after another and seeing how much - or indeed how little difference there is between them.
It also gives you average bitrate - for video only - and Q level across the entire file... once it loads in completely, which can take a couple minutes.
As an aside - if you can get away with CQ60, and everything still looks good, neither still scenes, smooth gradients nor action sections looking ropey, then I'd say stick with it. It's about the same level that a "normal" quality DVD recorder will capture in, and provides quite a good looking picture without excessive bitrate cost. (CBR VCD only looks bad at times because some - or all - of it goes up into the 10s, 20s... 30s... when stuck with an inefficient coder / matrix / settings against that 1150 maximum).
As stated, most scenes with single-figure Q levels should be perfectly acceptable, with increasing lack of 'clues' to it's compressed nature as the number falls - it's when it starts edging into double figures the trouble starts. And thats at some real low CQ values - 35-40 or less.
(the fact that the compression level tends to stay utterly constant also helps - it doesnt continually go from perfect to destroyed and back again, unless you have really challenging material... even a constant Q12 or so can look alright as your eye adapts to it ..... I've done Spirited Away (a semi-hand drawn animated feature, for those under rocks) ... all two hours-or-so of it...... on an 8cm CD. Mono sound and it took some trickery to get the DVD player to accept it, but it works and what's more, it's *watchable* ..... it was done with a CQ level of about 23, to-the-wire resizing (slight underscan), a mild soften filter and a max rate of about 960k. Some bits look about the same as the full size disc (very high action) but of course most parts look a lot rougher. However... your eye adapts! As that "haze"/distortion is always there! And after five minutes of rather challenging viewing, suddenly you don't notice it any more, like you would if it was coming and going)
And, um, yah. Still there? Good.
In summary: It's all to do with the kookiness in how CQ works. With "extreme" source material - very easily compressing, etc - it can be hard to shift the encoder off of the minimum bitrate (especially if it's quite conservative, rather than exploring what your player will allow) and minimum distortion without very large alterations.
And with some materials, the discernible difference between something taking 1000k to compress rather than 960k can be enormous.
something like that.
i'm still re-learning how to ride that particular bicycle....
((( but i'll have need of it soon, what with my R1 princess mononoke DVD, and now three more - recently released Nausicaa, Porco Rosso and The Cat Returns - on their way.... all itching to be turned into more palatable PAL, or at least R2 NTSC versions, for watching on the TV / 19" LCD monitor... agh.... so rusty.... how do you rip a wrong-region disc again? )))-= She sez there's ants in the carpet, dirty little monsters! =-
Back after a long time away, mainly because I now need to start making up vidcapped DVDRs for work and I haven't a clue where to start any more! -
wow, thanks for the info, just curious, how long did it take you to write that cuz it took me at least 30min to read and actually understand it, lol
----> adamf9898 ---->
Similar Threads
-
constant buzz from earphones
By ecos in forum Newbie / General discussionsReplies: 4Last Post: 4th Feb 2012, 18:43 -
Constant resolution in GK
By eliqush in forum EditingReplies: 0Last Post: 5th Aug 2009, 10:28 -
I need a constant bitrate
By alexv305 in forum Newbie / General discussionsReplies: 6Last Post: 6th May 2008, 06:38 -
Variable and Constant bitrate ??
By VEBouto in forum Newbie / General discussionsReplies: 4Last Post: 17th Oct 2007, 23:56 -
Staxrip: "Constant Quality" vs. "Exact File Size"
By FallenAngelII in forum Video ConversionReplies: 1Last Post: 9th Aug 2007, 17:40