Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    text
    copied!<p>Instead of chosing fixed bit rates, with the H.264 codec, you can also chose a different preset as described at <a href="https://trac.ffmpeg.org/wiki/x264EncodingGuide" rel="noreferrer">https://trac.ffmpeg.org/wiki/x264EncodingGuide</a>. I also found <a href="http://keyj.emphy.de/video-encoder-comparison/" rel="noreferrer">Video encoder comparison at KeyJ's blog</a> (<a href="http://archive.is/Tldgp" rel="noreferrer">archived version</a>) an interesting read, it compares H.264 against Theora and others.</p> <p>Following is a comparison of various options I tried. The recorded video was originally 673M in size, taken on an iPad using <a href="https://github.com/coolstar/RecordMyScreen" rel="noreferrer">RecordMyScreen</a>. It has a duration of about 20 minutes with a resolution of 1024x768 (with half of the video being blank, so I cropped it to 768x768). In order to reduce size, I lowered the resolution to 480x480. There is no audio.</p> <p>The results, taking the same 1024x768 as base (and applying cropping, scaling and <a href="https://stackoverflow.com/a/20313722/427545">a filter</a>):</p> <ul> <li>With no special options: 95M (encoding time: 1m19s).</li> <li>With only <code>-b 512k</code> added, the size dropped to 77M (encoding time: 1m17s).</li> <li>With only <code>-preset veryslow</code> (and no <code>-b</code>), it became 70M (encoding time: 6m14s)</li> <li>With both <code>-b 512k</code> and <code>-preset veryslow</code>, the size becomes 77M (100K smaller than just <code>-b 512k</code>).</li> <li>With <code>-preset veryslow -crf 28</code>, I get a file of 39M which took 5m47s (with no visual quality difference to me).</li> </ul> <p>N=1, so take the results with a grain of salt and perform your own tests.</p>
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload