by DJ_Izumi » Thu Jan 25, 2007 11:44 am
I... Think we may have been wrong here. While traditionally we say 'I frames only for editing', I actually put this to the test and have had supirising results.
I was initially just testing Lagarith agianst H.264 Lossless for compression for the purpose of storing masters. Not for editing.
For those interested, the souces was 60 seconds from the 5cm Per Second 720p trailer (I did this all at 720)
The Lagarith YV12 is 376mb (For 60 seconds of video)
The h.264 Lossless YV12 is 212mb. (Same segment of video)
Impressive savings, no? I noticed that the Lagarith used twice as much CPU speed as the h.264 Lossless for playback in Core Media Player. (Using CoreAVC)
So out of curiosity, I wanted to see how fast this h.264 seeks. I have 3 copies of VirtualDub open. One has the Lagarith opened directly. One has the h.264 Lossless loaded using DirectShowSource in an AVS script. The other has the original h.264 lossy encode of the trailer using the same AVS script.
Impressively, the Lagarith and the h.264 Lossless seem to seek equally as fast, infact very easily, and I sort of feel the h.264 may be seeking a bit FASTER and smoother than the Lagarith. I can drag the slider around and get my requested frames as I want them quickly and easily, and only with an Athlon 3200+ XP. The h.264 Lossy on the other hand, is slow as snot.
I think there may be actual merit in using h.264 Lossless as an editing codec, except for the problem of having to use AVISynth to get it into Premiere or something. At the least, I think this might be worth looking into.
For the sake of argument, our guides to tell people to use MPEG-2 streams via an AVISynth script to get them into Premiere, and MPEG-2 streams from DVDs arn'tt all I frames, are they?
I think I'm going to re-download Premiere (Recently reinstalled windows) and see what my results are.
