Encoding Advice Needed (Filesize Weirdness)

[Kevlar]

Hi everyone :)

I need some advice. Basically I've spent the last week or two searching the forums and experimenting with settings to encode 3 pr0n DVD's I own. My bias is towards quality over filesize, but I didn't want ridiculously huge files, so I went with:

Dideé's SixOfNine Max=20 CQM
1-Pass/fixed quant=3
640 x 480 res
AQ, QP, GMC and Trellis all on
VHQ=4 (and VHQ for b-frames)

Now I know from reading the forums that there's mixed opinion about whether turning on some of the above options helps or not in certain cases, but I was more than impressed with my final encodes so I left them on.

Or at least I was happy with the first two encodes. /:

DVD 1
Format: PAL
Length: 02:20:28
Encoded Size: 2.57GB
Bitrate (according to GSpot): 1892 kb/s
Encode Time: Approx. 13hrs

DVD 2
Format: PAL
Length: 02:19:50
Encoded Size: 2.08GB
Bitrate (according to GSpot): 1901 kb/s
Encode Time: Approx. 13hrs

I experimented with quant=4 for this one, and shaved a nice chunk off the filesize with a minimal quality loss. By minimal I mean if I was farther than 6in from the screen and not staring I couldn't notice it.

DVD 3
Format: NTSC
Length: 02:32:50
Predicted Encoded Size: 4GB +
Predicted Encode Time: Approx. 20hrs +

I got about 6 hours in before I aborted the 3rd encode (using quant=4 again btw). I couldn't see the point when the filesize would have been so huge. I might as well have just used DVD-Rebuilder to drop it to 4.5GB and left it at that.

Now I know that not all source material is created equal, and in fact the picture quality of the 3rd DVD was definitely better than the first two, but why the huge jump in filesize and encoding time? Do I just put it down to the fact that some material compresses better than others, choose a higher quant and move on? Or could it be something else?

Cheers,

-Kev

[Sirber]

Why not doing 2-pass VBR? :confused:

[Sharktooth]

1) Read the guides.
2) Re-read the guides.
3) Do it again.
4) Use 2-passes and set the final file size accordingly to your needs.
5) Use a lower bitrate matrix like 6of9 HVS (max=24...) or another one.
6) Check the compressibility and if it is lower than 60% reduce the resolution (resize) or add some compressibility-friendly filters (RemoveGrain(mode=2) or whatever helps compressibility without killing details) thru avisynth.

[Kevlar]

Er, maybe I didn't make my original post clear enough. I tend to babble on when explaining things so it wouldn't surprise me. :D

I wasn't asking how I should encode the movie. I've encoded many movies in the past. I chose fixed quant encodes because I wanted to keep a similar level of quality across the 3 movies.

What's puzzling me is how the first two could be so similar in terms of encoding time and filesize, yet the third goes through the roof. The third DVD was definitely of better quality, but enough so that the final encode takes 7 hours longer and weighs in 2GB larger? That's what I don't get.

So what I'm asking is, is it "normal" to see such large jumps in encoding time and final filesize when doing fixed quant encodings, even with (reasonably) similar material?

Sharktooth: I'll try Dideé's other matrix...forgot about that one. And yes, I've read the guides. :rolleyes:

[Sharktooth]

The "jump thru the roof" is due to a lower overall compressibility of the 3rd source.
There are different solutions to fix that size "problem" (it's perfectly normal though): using a lower resolution, a lower bitrate matrix, avisynth filters to rise compressibility (denoising, smoothing etc.), lower the framerate (not as good as the previous choices...), etc.

[darkavatar1470]

well, it can happen.

Sometimes during the first half of encoding, I get the projected file size & encoding time blown up to 4x the actual vaule, but it will zoom on to the correct time & size after a few hours.

So maybe you should wait for it to finish, and if its still oversized, you can check the stream to see what happened¢wmaybe it's encoding all I,P frames, and the quants are locked to 2~4 or something...

[Didée]

Originally posted by Kevlar
What's puzzling me is how the first two could be so similar in terms of encoding time and filesize, yet the third goes through the roof.

***

DVD 1
Format: PAL

DVD 2
Format: PAL

DVD 3
Format: NTSC

See a relation here?

If the NTSC source were telecined or interlaced, and you had done a straight encoding without IVTC or deinterlacing, it would perfectly fit the picture ;)

[Sharktooth]

Ah! Didnt spot it :eek:

[Kevlar]

Ahh...so performing IVTC can cause the symptoms I stated? Or rather, that and the fact that the third source was of better quality combined would cause the symptoms? I had to deinterlace all three, so the encodes were no different on that front.

If so, that's cool. Just trying to understand why things were so different. I was sure I'd done something wrong...

P.S. I tried SixOfNine-HVS @ quant=5 and let it run for a bit, and it predicted filesize was teetering between 3.2 - 3.5GB. It looked pretty damn good, too. I think I might give it another whirl @ quant=4 and let if run all the way.

Cheers guys. :thanks: