yaz
5th March 2004, 13:05
afais, lot of pp use this cute little stuff. at the moment i'm not aware of any other way to have a look inside devapi4 xvids. i haven't got any problem with it ... so far.
having analyzed my last encodes i found sg strange. when more than 1 conseq b-frame is set together with gmc (s-frames) drfa reports some occasional jerks in drf usage. it states that time to time quants jump up to 31 producing a small bounch of such frames & then everything goes as expected. the percentage of such frames for a whole film is far below 1%. meanwhile, the average quantizer for the film seems to be ok.
when analyzed with ffdshow, no such qaunt jerks was found.
when quants were restricted (i never do, anyway) the results were the same.
when only 1 conseq b-frame's allowed or when gmc's off, no jerk anywhere. that's why i think that b & s-frames together(!) cause this.
yesterday night i went on a frame by frame exam & i found that the first occurances of these scrambles (i went through only on some of them) were sw round where some conseq i-frames(!?!?) were also present. so ... i'm completely confused now. can anyone explain me this ?
some quests just occured to me about it.
- how safe & reliable is gmc now. i know, i know ... it is(!:-) but anyway. what risk should i take into account (if any) ?
- how safe & reliable is using B & S frames together ? so far i've encoded 8 movies with this settings & i found that b & s frames take more than 60% of all but st it goes far above 70%!
sg ot.
yesterday night i tried the 03040900 build from gam3r & all of my encodes mentioned earlier happened to be decoded as a crap. i've seen from the logbook that sg's been changed in the 'gmc/motion' part. is there sg about to change in it essentially ? if yes, what ? (with older builds decoding seems to be ok)
thx
y
having analyzed my last encodes i found sg strange. when more than 1 conseq b-frame is set together with gmc (s-frames) drfa reports some occasional jerks in drf usage. it states that time to time quants jump up to 31 producing a small bounch of such frames & then everything goes as expected. the percentage of such frames for a whole film is far below 1%. meanwhile, the average quantizer for the film seems to be ok.
when analyzed with ffdshow, no such qaunt jerks was found.
when quants were restricted (i never do, anyway) the results were the same.
when only 1 conseq b-frame's allowed or when gmc's off, no jerk anywhere. that's why i think that b & s-frames together(!) cause this.
yesterday night i went on a frame by frame exam & i found that the first occurances of these scrambles (i went through only on some of them) were sw round where some conseq i-frames(!?!?) were also present. so ... i'm completely confused now. can anyone explain me this ?
some quests just occured to me about it.
- how safe & reliable is gmc now. i know, i know ... it is(!:-) but anyway. what risk should i take into account (if any) ?
- how safe & reliable is using B & S frames together ? so far i've encoded 8 movies with this settings & i found that b & s frames take more than 60% of all but st it goes far above 70%!
sg ot.
yesterday night i tried the 03040900 build from gam3r & all of my encodes mentioned earlier happened to be decoded as a crap. i've seen from the logbook that sg's been changed in the 'gmc/motion' part. is there sg about to change in it essentially ? if yes, what ? (with older builds decoding seems to be ok)
thx
y