When encoding a video which has a significant section that is completely black with ffmpeg mpeg4, the 3-pass encoder pushes up the quantizers to 31 in the black region. This only happens in the third pass for some reason. The problem is that it does not really save any bits, and when some more complex content follows the black region, it takes a few seconds until the quantizers are reduced back to sane levels, and the video looks really horrible during that time. I have tried to set vqblur=0.0 and I've tried to use vrc_override, but neither helps. It seems that vrc_override is completely ignored in the third pass, and vqblur worked exactly opposite as I expected, after setting vqblur=0.0 it took a longer to reach q=31 in the blank region, but there was no change in the rate of decrease of the quantization levels after the blank region. One solution would be is to avoid pushing up the quantization at blank frames, but if an automatic solution is difficult, respecting vrc_override would help too. Thanks, Zoli