I am trying to profile the affect of movement on bitrate and
therefore the volume of data required to stream flash h.264 video.
The reason why I am doing this is because the studio I am
setting up will have a number of LCD video screens behind the
presenter on camera. This wall of LCD screens will be displaying
different TV feeds. To simulate and analyze the affect of this on
stream size - I have setup 6 normal pc lcd screens and pointed a
webcam at it. Encoding this video for 1 minute intervals in FME I
am comparing the output file sizes after one minute of recording. I
started off with no LCD screens displaying anything and worked my
way up to having all 6 screens on. I am using the default profile -
"Medium Bandwidth H.264 (300kbps)".
My expectation was that with no screens on and therefore no
movement other than changes in ambient light - that the file size
would be dramatically different than with all the lcd screens
showing different tv feeds. This expectation has come out of my
research that says when streaming h264 - pframes are sent which
include only sections that have changed – as opposed to
iframes which have the full frame.
However – when I compare my resulting output files
there is no perceived difference in the output file sizes.
Is there some fundamental issue with my testing or my
expectation that a stream with a higher bitrates will result in
much larger files – given a constant window of time?
Am I missing some settings in FME?
On an average the bitrate will remain constant ( 300kbps in
your case) for H.264 irrespective of the quality of input video. If
the device sends less/more frames than the agreed fps, then the
bitrate will decrease/increase accordingly.
But with vp6, the input video quality may impact the
I have another problem: Quality of h.264 video doesn't seem
to change, even if we alter the bitrate quite dramastic.
Eg. 640x480 video looks quite the same, whether encoded with
350kbit or with 1000 kbit.
That can't be right, especially because the quality is always
It is very unlikely that the video quality would be same on
350 & 1000kbps.
1. What is the output bitrate listed in the FME Encoding Log
2. What is the CPU utilization? If it's greater than 70%,
then you may need higher end machines.