(The actual numbers vary depending on the source geometry, but these are good numbers for "typical" source)
Aspect ratio compensation for widescreen (> 16:9 asr) content encoded to --export_asr 3:
Note that when you are encoding content that is 16:9 or wider, you should tell transcode that it is encoded at 16:9, this will ensure that the geometry calculations for export are correct (do this even if the source was originally letterboxed into 4:3 and subsequently cropped):
Denoising with moderate sharpening should always be used with anime (gets rid of the "ringing" around subtitles and other sharp lines):
A gamma boost is almost always needed as well:
See the eq2 section of the man page for mplayer:
-x mplayer="-vf eq2=NN:NN:NN:NN:NN:NN:NN:NN"
eq2[=gamma:contrast:brightness:saturation:rg:gg:bg:weight] Alternative software equalizer that uses lookup tables (very slow), allowing gamma correction in addition to simple bright- ness and contrast adjustment. Note that it uses the same MMX optimized code as -vf eq if all gamma values are 1.0. The pa- rameters are given as floating point values. <0.1-10> initial gamma value (default: 1.0) <-2-2> initial contrast, where negative values result in a neg- ative image (default: 1.0) <-1-1> initial brightness (default: 0.0) <0-3> initial saturation (default: 1.0) <0.1-10> gamma value for the red component (default: 1.0) <0.1-10> gamma value for the green component (default: 1.0) <0.1-10> gamma value for the blue component (default: 1.0) <0-1> The weight parameter can be used to reduce the effect of a high gamma value on bright image areas, e.g. keep them from getting overamplified and just plain white. A val- ue of 0.0 turns the gamma correction all the way down while 1.0 leaves it at its full strength (default: 1.0).
Good starting values for a typical slightly flat source: -vf eq2=0.9:1:0.05:1.03
For a VHS rip with red bloom: -vf eq2=0.9:1.1:-0.05:1:0.9:1:1,hqdn3d=3:2:3:3
Using this ./ffmpeg.cfg file should kick ffmpeg up to mpeg2enc quality:
[mpeg2video] mbd = 2 mv0 = 1 nr = 20 trell = 1 vqcomp = 0.7 vqblur = 0.3
And this one should make the output indistinguishable from the input, at a HUGE cost in encoding time:
[mpeg2video] mbd = 2 mv0 = 1 nr = 20 trell = 1 cmp = 3 precmp = 3 subcmp = 3 vqcomp = 0.7 vqblur = 0.3