[x264-devel] VBV single frame prediction

[Paul McDermott]

> My guess is that your video is very low bitrate/low resolution, which makes it a lot more difficult to achieve exact accuracy

The video I posted in the annexb format might have been both low bitrate and low resolution, but with the same configuration for VBV, videos of all bitrates/resolutions exhibit similar output from the encoder. 

Whilst individual frames can vary by as much  5-50% less than the target output size, overall average across videos is ~11% that is required to be padded in the buffer.

Making a significantly smaller buffer size (e.g. 20% less) across all frames and re-encoding any frames which overflow at a lower bitrate in VBV is something I'm considering, in an attempt to recover as much of this 11% padding as possible.

Thanks,
Paul