View Full Version : Increasing Filesize and Pass 2 Questions
Asrial
2nd September 2003, 23:15
1) My understanding on pass 2 encoding is that the 2nd pass takes information from pass 1 and the only thing it does to reach the desired filesize is that it lowers the bitrate in areas it can get away with it at. Is this correct? If so, that means that it's better to reduce filesize with B-frames and R-D Trellis and the rest rather than rely on pass 2 removing bitrate, right?
2) If pass 1 is already at the filesize I want, then it's better to go with pass 1 over pass 2, right? However, if question 1 is correct then pass 2 is just wasting your time because it's not really doing anything. Right?
3) I'm wanting an encode that's 200mb but it's coming out too small. Is there an easy method to determine what to do to get it to be 200mb or does it just involve trial and error? Furthermore, would lowering B-frames, unchecking R-D Trellis, setting VHQ to 1, etc. be more harmful than helpful?
I consider myself to have an average understanding of media encoding but still consider myself a novice with regards to XviD.
Danzel
3rd September 2003, 00:17
1) 2 pass is designed to allow you to hit a required file size by scaling down the bitrate used. Although adding more B frames and such will give you a lower file size, 2 pass will give you more accurate power over hitting your required file size, where extra B frames may end up undersizing it.
2) if your pass 1 is already at the size you want and it looks fine then you dont need to do a pass 2, however you could lower the B frames and do your 2 passes again, as B frames are more processor intensive to decode than P frames.
3) If you want to hit a file size and its coming out too small, things you can try are:
Increase your video resoloution.
Use a sharper Resizer.
Decrease your B frames.
Use Mpeg Quant instead of H263.
(Theres more things you could do, but thats a few)
Im not sure if Trellis gives a smaller file size, but it does make things look better (to my eyes).
Lowering your B Frames will leave you with a bigger file.
Higher VHQ gives a higher PSNR. (check out this post (http://forum.doom9.org/showthread.php?s=&threadid=58606&highlight=vhq) )
I usually encode my first pass with my usual setting and if it comes out to small then ill increase my resolution/decrease my B frames so that my first pass comes out bigger than the desired size.
Danzel.
PS. Welcome to the Forums.
Asrial
3rd September 2003, 01:28
Originally posted by Danzel
PS. Welcome to the Forums. Thanks! :)
I don't know a thing about PSNR but that's because I haven't found the 'newbie' guide for it yet.
sysKin
3rd September 2003, 14:10
Originally posted by Asrial
1) My understanding on pass 2 encoding is that the 2nd pass takes information from pass 1 and the only thing it does to reach the desired filesize is that it lowers the bitrate in areas it can get away with it at. Is this correct?Actually it just scales down the whole bitrate, assuming there is no such place where it would get away with it more than somewhere else (and it's mostly true, with some exceptions which are my ideas for post-1.0 development).
If so, that means that it's better to reduce filesize with B-frames and R-D Trellis and the rest rather than rely on pass 2 removing bitrate, right?Second pass is the way to do it. 2pass code aims for given bitrate while trellis, b-frames etc ensure maximum quality at this bitrate (in theory of course, but it's still true in practice).
2) If pass 1 is already at the filesize I want, then it's better to go with pass 1 over pass 2, right? However, if question 1 is correct then pass 2 is just wasting your time because it's not really doing anything. Right?Correct. If you suspect that first pass can give you the filesize you want, disable "discard first pass" in configuration - you'll have the file ready after first pass (or you can still do 2nd pass if needed).
3) I'm wanting an encode that's 200mb but it's coming out too small. Is there an easy method to determine what to do to get it to be 200mb or does it just involve trial and error? Furthermore, would lowering B-frames, unchecking R-D Trellis, setting VHQ to 1, etc. be more harmful than helpful?The best way is to use all the fancy options, but change quantizer matrix to high-quality one (HVS-best and such like). Disabling b-frames will also improve quality (but not as well), same for trellis. Lowering VHQ will *lower* quality - which doesn't mean that you'll notice a difference ;)
Radek
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.