Is frameserving really always better?

[Solarjetman]

I am new to video editing (been doing it for about 3 weeks now) and have a question about how useful frameserving actually is. I have been doing my homework on the subject, and many people seem to believe that you should always frameserve. However, aren't there times where this would actually be a bad idea?

From what I understand, a frameserver takes a single frame from a video, passes it through the applied filters, then gives the frame to the encoder.

However, if the encoder is running in 2-pass VBR mode, wouldn't you be running the same frame through your filters twice (once for the first and agian for the second pass)? To me it would seem that if you had the disk space, it would be faster to create an AVI file, then run the AVI file through the encoder.

Also, by creating the AVI file, you could apply temporal filters which you couldn't while frameserving.

Am I missing something here? Thanks all.

[nick1977]

In my experience, it's just more convenient than outputting to a new avi file. Also, you will usually need loads of additional space free unless you use compression. Using an additional stage of compression will reduce the eventual output quality anyway.
On top of that, unless you are using *loads* of filters in virtualdub, it doesn't seem to add too much to the encoding time. Well, not enough for me to output to an avi first.

I guess it's just a matter of preference, but for me the advantages outweigh the only disadvantage i.e. filters being applied twice.

Also, I don't see why you can't use temporal filters when frameserving. I use temporal smoother quite regularly when frameserving.

Nick

[Solarjetman]

Also, I don't see why you can't use temporal filters when frameserving. I use temporal smoother quite regularly when frameserving.

Thanks Nick. For the most part I agree with you. But from the VirtualDub website they say this...

Which options do not work during frameserving?

The following options will work:

Video filters.
Audio offset (displacement/skew correction).
Audio source switching (remultiplexing).
Range cropping (start/end) and segment deletion (i.e. crappy edits).

The following options will not:

Inverse telecine (IVTC/3:2 pulldown removal).
Audio conversion.
Audio interleaving -- this is handled by the client app.
Audio and video compression.
Any filter that shows a lag value in its description, particularly the "temporal smoother."

Do you actually see a difference with the temporal smoother? It does name that specificly as one that would not work. I have not tried it though.

[nick1977]

Very odd that it lists temporal filter as one that doesn't work. I can assure you it definitely does for me! I've been encoding some old cartoons from 1980's VHS. Naturally the source is very noisy, so the temporal smoother does a lovely job of "cleaning" the image, as well as helping to stabalise it. It is probably one of the filters I use most frequently to be honest.

I would say try it, and see if it works for you. If it doesn't I'll let you know my version numbers of virtualdub and tmpgenc.

Who knows maybe I'm just lucky

Nick

[rob1]

If you have the hard drive space and are running +3 passes I've found it's faster to filter and save as an avi and THEN encode than it is to frameserve (though I use avisynth, should be same difference).

For example, it generally takes me around .5 real time PER PASS to encode using my basic filter set and frameserving to CCE. If I instead frameserve to vdub and save (using huffy) to an avi, that avi will encode in real time (per pass). BUT, it takes a bloody long time to save the filtered avi. I actually tried it both ways and I can't remember the numbers, but for me, if I'm doing 3/4 passes, it's faster to save as an avi and then encode. The more passes, the greater the difference. For two pass encoding, frameserving was faster. Does that make sense?

That said, I usually encode when I go to bed and so I generally just frameserve it anyway- and no problems with temporal smoothers and frameserving via avisynth.

[SLK001]

All the decoder does on all but the last pass is update the .VAF file - it doesn't encode the video (just yet - but it DOES examine the video very closely). So any filtering that you do will be presented to the decoder exactly the same way on each pass. On the last pass, the .VAF file is used by the decoder to allocate BR, scene breaks and the like, AND everything is written to HD.

[Solarjetman]

I would say try it, and see if it works for you. If it doesn't I'll let you know my version numbers of virtualdub and tmpgenc.

Who knows maybe I'm just lucky

Nick

Well, you are lucky. Vdub gives me a warning when starting the frameserver that the audio will be out of sync because a filter has a lag field. I forgot the exact message but that is the gist of it. I did not bother trying to go through with it though to see the effects. I am using version 1.5.4

[disturbed1]

Save the audio file as a wav.

Frame serve the video, and load the wav file seperate, encode together.

[MrKGB]

Save the audio file as a wav.

Frame serve the video, and load the wav file seperate, encode together.

Too complicated...

Just compensate the lag: In vdub Audio->Audio Video Int Opt->Delay... put something like -300 (don't remeber exact value) and frameserve.

Or even easier: use TemporalSmoother or Convolution3d of AviSynth.

[ZippyP.]

Too complicated...

Not really, and it avoids a/v sync problems. Extract a wav file, use that as the audio source. Simple.

[wulf109]

If your source is a DVD then TMPEG should be able to encode directly from the vob file. I always did my SVCD's this way.