Andrew Church wrote: > >Can anybody comment on *why* TMPGEnc produces higher quality > >output? Is it just well-chosen defaults? I find that I can > >do pretty nearly whatever I want in terms of visual quality > >by using lavc options and filter plugins. > > I don't know, but I've observed the same myself (see > http://www.transcoding.org/cgi-bin/transcode?FFmpeg_Vs._Mpeg2enc). > I'd have to dissect the TMpgEnc binary to figure out more, but given > that it takes significantly longer than either ffmpeg or mpeg2enc to > encode a given video clip, my completely uneducated guess is that it > tries many more parameter variations during the encoding process than > ffmpeg and mpeg2enc do.
So it is probably possible to achieve the same results using some combination of lavc options. In my experience, agressive use of hqdn3d will do more for animated material than any amount of other fiddling, and that a little gamma boost (0.9) will improve almost anything when targeting dvd. Also, using mbd=2, v4mv and trell seem to encourage better use of bits. And of course, the difference between single and two-pass is night-and-day. At this point I am entirely happy with all my encodes, especially since the advent of yadif. I guess I'm just a crank for refusing to give any credit to any software that targets windows exclusively, but when I discovered that it was a much simpler matter to port DOS code to Linux and X than to windows (in 1995) I made a decision that I am not going to back out of. Perhaps we could start providing encoding profiles for different types of source and target on the wiki, to even up the results of any future shoot-outs?
