Fast recompress for the first pass?

[MikeHunt79]

Hi, my source for compressing to XVID is not DVD, I'm trying to compress things that I've captured using my All-in-wonder Radeon card. (MPEG2 files,basically).

These MPEG-2's have often have noise in the picture, so I use VirtualDub's exellent Temporal Smoothing filter to reduce or eliminate this noise, The only problem with this filter is that it seems to halve the encoding speed of my machine.

I'm also using a Resize filter aswell, to reduce the reolution to 480 x 360.

So... Doom9's Xvid guide says to to "Fast recompress" for the first pass, yet Vdub gives a Warning that the filters are being bypassed. This means that the bitrate curve is being based on the noisy video. Does this mean the results will be much worse than using "Full processing mode", where the filters are used to produce the first pass?

[valipod]

In my experience there are only pluses when using fast recompress (colorspace ->> speed etc.), but you have to use filters outside VirtualDub, aviSynth that is. Not only because this is the only way to use filters with fast recompress, but they are far better than the ones built in Virtual Dub.

Try Gordian Knot for generating an avisynth script, you'll see it's very simple.

[EthanoliX]

From my experience I would not recommend deactivating Filters during first pass ("fast recompress"). Encoding might be much faster, but I think XviD makes the wrong I/P/B-Frame decision, when filters are only activated in second pass. Especial if you're doing a strong filtering the source XviD gets in both passes is very different from each other.

[Danzel]

Definately use the same filters during each pass, otherwise Xvid will be chosing the wrong frame types and sizes during the 2nd pass.

This means that you should either use VDubs Filters and full recompress on both passes (This has extra color space conversions, which is bad)

Or, Use AviSynth to do your noise removal and resizing, which will allow you to use Fast Recompress and Generally this method will be quicker, and with no extra colorspace conversions, should give better results.

Check the AviSynth forum for more information.
Or give Gordian Knots AviSynth Script making functions a go.

Danzel.

[MikeHunt79]

Thanx for the help...

I've tried various experiments (filters on and off on the first pass), and eventuially I got the gordian knot method to work, and it is much faster than Vdub, even with the Denoise filters that gordian/avisynth has turned on for the first pass. I'm now getting 30fps at times!

I'm now sticking with this method, even tho DVD2AVI takes quite a while to process all those MPEG2's.

[mikeX]

if i'm not mistaken, this little plugin (http://nic.dnsalias.com/mpeg2dec3.html) can handle mpeg2 files directly, without the need of dvd2avi project files

[MikeHunt79]

mikeX: I tried it out, but Gknot wouldn't import the mpeg directly... I even tried editing the .avs file manually, but with no joy. I don't really mind running the file thru DVD2AVI really, since Vdub takes a similar amount of time to import the mpeg directly anyway.

[mikeX]

oops, i actually meant this little plugin (http://nic.dnsalias.com/mpegdecoder.html)

(you can't use it with GKnot)


but still i can't remember if it can actually bypass d2v projects or if it's even faster...