[karl_lillevold]

@superdump: the difference in speed is due to the (previous) difference between 80 and 85. In order to get quality 85 with M6, you now have to use 90. However, the improvement in quality from 80 to 90 is not worth the extra time for now.

Yes, RV9 has an adaptive advanced in-loop filter, specifically designed to remove block edges before using the reconstructed frame in the prediction loop. Therefore the encoder and decoder filters have to match 100%, and the filtering strength can not be adjusted on the decoder side. It is not the same filter as in H.264, but the principle is the same in both, if you are looking for technical background on the subject. Specifically, one feature RealNetworks proposed to be included in H.264, was a tiny amount of pre-determined dither across filtered block boundaries, because on large (LCD, plasma, DLP) screens, a difference of just '1' in luminance or chrominance level is visible. However, we did not succeed, and the H.264 in-loop filter does not have this feature.

@alx: thanks! that's a very good suggestion. We have already considered this option, and it may be included in Producer 9.3, but not in 9.2, since it requires quite a few producer changes. One problem is that a current requirement is that the two passes are run with the exact same parameters, including bitrate, resizers, pre-filters. If any of those are accidentally changed, the 2nd pass would be messed up, usually resulting in too small filesize.