Fastest Encoding from a VOB?

What iDCT? What program? What codec (if not xvid) I have an Athlon XP, and i want te achieve speeds higher than 20 FPS (i only get these from a VOB-converting from a MPEG is @ 50)

Vidomi is THE fastest one. With divx5.02 b-frames and mp3 transcoding at the same time. 17 fps on 900Mhz celeron.

oh god, please don't recommend vidomi. And what's your rush for speed? I encode at about 2-4 FPS on a 1.3ghz duron with lots of filters and a pretty large resolution. If you want something easy, I'd familiarize yourself with autodub 1.7 final release. Then if you want more power, you can manipulate the individual programs/components that make up autodub.

I had no problems with Vidomi on the video part, but the AC3->MP3 muxing really sucked :rolleyes: . If you would want raw speed and don't mind geeking around in Linux, I suggest mencoder. Bilu

Nics MPEGDecoder.dll also gives a little Speedboost :)

Use 1st pass for 2nd pass 

Is it possible to load a VOB file with VDubMod / AviSynth, run XVID first pass on it and do not discard the video stream but use it as source for the second pass ? This way the AviSynth filters would not have to be processed twice. If the quality of the 1st pass is too bad, would it be possible to add a quality slider for 1st pass (like the 1pass quantizer) ? Thx

Nightscape, The most obvious way of doing that would be to run the video through your filters into a HUFFYUV file, then use that as the source for both passes. Of course that takes up a lot of space, but it is truly temporary as you can delete the VOB as soon as you have the HUFFYUV file verified.

For me, this is hardly practicable. A complete DVD would take several GBytes on hard disk. And XVid does the encoding anyway, so why not in a way that I can use for further processing ?

XviD does have an MV Hinting function (unless it isn't working), but this does reduce the quality of the second pass. The reason compression has to be duplicated is that the second pass isn't a perfect clone of the first: Different quantizers and settings are used, meaning that you can't just copy data in bulk from the first pass and have it work right.