Video sizes with lossless and lossy codecs

Everything not related to Prince of Persia should be discussed here.
Post Reply
David
The Prince of Persia
The Prince of Persia
Posts: 2846
Joined: December 11th, 2008, 9:48 pm
Location: Hungary

Video sizes with lossless and lossy codecs

Post by David »

Why is it that PoP videos are usually bigger with the lossy OGG Theora codec than with the lossless ZMBV codec?
Maybe it's because lossy codecs are designed for smooth transitions, which PoP videos don't have?

For example, look at the videos here: viewtopic.php?p=12715#p12715

Code: Select all

                        ZMBV (bytes) OGG Theora
backstab2                66 230   292 226
guard_gone               45 252   164 110
guard_survives_chomper   71 592   343 318
guard_wrap               69 174   193 841
immortaldrop_guard2      69 490   266 418
spike_2                  70 380   193 399
tapestry_disables_guard  45 270   138 442
throughfloor_opendoor   130 364   590 423
User avatar
Norbert
The Prince of Persia
The Prince of Persia
Posts: 5743
Joined: April 9th, 2009, 10:58 pm

Re: Video sizes with lossless and lossy codecs

Post by Norbert »

David wrote:Why is it that PoP videos are usually bigger with the lossy OGG Theora codec than with the lossless ZMBV codec?
I'm not familiar enough with the subject matter to give you an insightful answer, but here are my thoughts. The lossless stream is a recording, while the lossy stream comes from the output of a converter. This means you need to take into account at least these two things: 1. if the game footage was directly captured to Theora, the file size might be smaller, and 2. I used "-v 10" video quality when converting with ffmpeg2theora, which is the highest encoding quality (the default is 6) and presumably makes the output as non-lossy as possible. Also, it does seem that ZMBV (heavily?) uses previous frames, and lossless PoP1 footage includes almost no pixels that change. Even moving to another room is probably just one frame, which is usually still not a full redraw given the similarity of many rooms and tiles. Theora's algorithm discrete cosine transform may be less suitable to process such footage. Just some speculation.
David
The Prince of Persia
The Prince of Persia
Posts: 2846
Joined: December 11th, 2008, 9:48 pm
Location: Hungary

Re: Video sizes with lossless and lossy codecs

Post by David »

Norbert wrote:Also, it does seem that ZMBV (heavily?) uses previous frames
Theora has inter frames (P-Frames) which is pretty much the same thing.
Post Reply