Filters and speed reduction, a question

[apfraats]

As I 've metioned in my thread NOISE, IS IT JUST ME ?, I'm now convinced is just because of bad source material. I just used DVD-REMAKE-PRO to cut down 50% from the DVD-movie, so one should have a perfect 1:1 copy of the original, I suppose, concerning the MPEG-stream.

Still the noise was there , even in the original, and that's refering to the 'snowy noise'. On the introducing studio logo's the GIBB-artifacts are already slightly seen, and if you recode everything at a lower average bitrate they just get a bit worse.....

But I've tried the filters UNDOT().DEEN() now, and the result is much better, concerning the snowy noice.

I was already warned by JDOBBS that the use of filtering takes extra encoding time, but I noticed it takes extra encoding time in ALL PASSES being done....

That I don't understand. I was just thinking the filter was applied against the output of AVISYNTH, so I did expect the VAF creation to be slower....

But also the MPEG creation is much slower now, and I do three MPEG passes and even the speed on the second and third pass is down....

I wouldn't expect that. I would understand if the VAF creation was slower, or the first MPEG encoding pass, but not the repeated VBR MPEG ecncoding passes.


All passes are slowed down to 1.00. So it takes whole day now to do a movie........

Can anybody explain this to me ???

Thanks in advance..

[jdobbs]

Each pass reads the original again, not just the VAF pass. In fact the functioning is the same for any of the passes, even the VAF one... the only difference is whether you have information from the previous pass that can help determine bit allocation.

[apfraats]

Each pass reads the original again, not just the VAF pass. In fact the functioning is the same for any of the passes, even the VAF one... the only difference is whether you have information from the previous pass that can help determine bit allocation.


So this is true, you are the one knowing it. If it's just about bitrate allocation, I suppose the use of doing x passes over x-1 passes (x-1)>0 has no use alltogehether ??? I thought it improved quality by using the already encoded stream info and again looking at the source and improve the Mpeg stream. This is not what happes though, only you'll get better bitrate allocation.

So the use of more than 2 not-VAF passes is verry limitted I understand ???

If the filters are applied every-time, why isn't there an option to use the filters once against the source and temorarely holding these files untill enosing has finished. In this way a bit more space is needed, but harddisk are giant already nowdays.
It however will save significant time when you do multiple pass encoding and are using a filter.....

So maybe an option 'Save Temporary AVISYNTH files' or something like that would be handy for all users using filters ????

It will greatly save time on multipass VBR encoding...

[jdobbs]

That's the reason the improvement gets smaller with each pass. In fact after the 2nd it is exceptionally small, 3rd smaller, and after the 4th it isn't detectable any more.