Does 2-Pass Really work?

[bobby8798]

I encoded a clip using 1 Pass CBR

Then with the identical avs input file, I did a 2 Pass encode.
On the 2nd pass, I used the same target filesize as from the 1-pass encode

I expected the 2-pass to be better because it SHOULD allocate more bits to the scenes that need them, and take bits from scenes that don't need as many bits.

I compared the the 1-pass and 2-pass clips Frame-by-Frame, and they were IDENTICAL.

The clip is 15mins long.

Has anyone ever tried this to confirm if 2-pass is really better?
I am surprised by the results.

using XviD-02022003-1.exe Koepi

[WyldeF1r3]

I bet their sizes weren't identical though ^_~

CBR will overdistribute and understribute bits to various scenes. VBR (2-pass) will not; unless you give it way too much bitrate). The point of 2pass is best quality at smallest size. A 2pass and a 1pass quality based may have the exact same quality (to the naked eye), but the 2pass one will most likely be a good deal smaller than the 1pass encoding. The same holds true for CBR. Also, a great deal of the time it depends on the source you're encoding. Some are incredibly compressable, and some are really not. And of course, if you both set those output filesizes large it wouldn't matter anyway. The way to tell the true advantage is at lower bitrates and higher quants.

[McQuaid]

Is there a way to determine what size the second pass should be will retaining the quality of the 1st pass?

When people encode movies to a cd this is not an issue cause you just make it to 700 meg size. But I am encoding tv shows, so say a 20 min show comes to 183megs in the first pass, is there a way to know the exact value that will retain the first pass quality while being the smallest file size in doing so?

[bobby8798]

1-pass file = 221,530 KB

2-pass file = 221,540 KB

Bit Rate = 2000

[TelemachusMH]

Try lowering the bitrate to 600 kbps and try your test again ... the results should be much clearer then.

TelemachusMH

[WyldeF1r3]

Originally posted by McQuaid
Is there a way to determine what size the second pass should be will retaining the quality of the 1st pass?

run your 1st pass then open gknot and click on the "nandub files" tab and open your stats file. It will open an xvid stats file. Then click on the "bitrate" tab. Set the file size you think the episode should be then look at the "Video Size/First Pass" box (slightly down and to the left of where you just input your file size). There you will see a % value. You want to set your file size so that value = as close to 60% as possible. That will inform you what the smallest size you can have is while maintaining as much quality as possible.

[jarthel]

Originally posted by bobby8798
1-pass file = 221,530 KB

2-pass file = 221,540 KB

Bit Rate = 2000

2000 kbps? So that's why your 1st pass CBR quality is so good.

[bobby8798]

If 2-Pass works, it shouldn't matter if the bitrate is 2000

There is still room to increase or reduce by a quantizer, saving bits to improve quality for the frames that need it.

Maybe I am doing something wrong.

I stumbled upon a thread where someone said 2-pass on certain builds doesn't work, but I can't find it now.

[tangent]

The point is, at 2000kbps, almost all your frames will be at quantiser 2, so there's hardly little room or need to increase or lower quantisers. If you want to make a comparison, do it at a lower bitrate.

[bobby8798]

Test with Quantizer = 2 FIXED

File size is OVER 500M

Has anyone actually tried a test to determine if 2-pass is really effective, and have better quality without having to LOWER the bitrate?

It would be nice to encode with Quantizer=2, but the file size would be too large.

[Selur]

Could you at least post some more info's about your settings, your source material, logfiles, tools & scripts you use etc.
see: http://forum.doom9.org/showthread.php?s=&threadid=24924

Did you try if your problem presists with a newer build?