Re: [libreoffice-marketing] Re: [ACTION ITEM] LibreOffice Video links

Hi Marc, *,

On Sun, Mar 10, 2013 at 3:50 AM, Marc Paré <marc@marcpare.com> wrote:
> Le 09/03/13 06:46 PM, Christian Lohmaier a écrit :
> [...]
> Speaking of nl groups. I know that video production is one of your interests
> ... could you also mention the type of protocols we could use for video
> production that would make the job of translation of videos by other nl
> teams? What software would you recommend?

As youtube doesn't allow multiple audio tracks, the only way to offer
different languages via subtitles (or upload the same video with the
audio track replaced as a separate video).

I personally do create subtitles/transcripts with aegisub, as it has
nice timing features - I also mainly use it because i can use
formatting and positioning that I use in the credit page. Also the
fade-to-black on my video(s) or the pulsing fast-forward marker are
done with styled subtitles. But all that is not required for
translation, as youtube only can display unformatted ones. So any
subtitleeditor you can find out there will work.

I don't have a checklist, but some general guidelines:

* never create a silent video (unless it is very, very short). Always
either add a spoken commentary/explanation or some background music.
Watching completely silent video is alienating, and many people will
not watch the whole thing.

* if you're doing a screencast, record it in a lossless format. 15fps
is enough unless you have lots of motion in your video (but that
typically is not the case for screencasts) [1]

* if you're doing a screencast, don't record in 1080p with tiny fonts.
Think of those who will display the video in 720p or in 480p. I.e. use
a comfortable screen size.

* If you're recording regular video, and intend to point your camera
to a screen, make sure that your room is well-lit, otherwise the
contrast between the bright (light-emitting) screen and the room is
too much for the camera's brightness control. Consider preparing a
screencast of the action that is going on on-screen and edit your
video later and mix life-capture and the on-screen stuff. (using two
separate video cameras might also be an option, if available)

* When recording audio, thing beforehand about what you want to talk
about (or edit your video later and rerecord the overdub later). If
there are too much pauses because of "what should I do or say next" it
gets annoying very fast.

* When recording audio, make sure to check your recording level prior
to recording the whole thing. Nothing more annoying to have clipped
audio or too faint volume throughout the whole thing. If the position
of the microphone is fixed relative to the speaker, consider locking
the input level as opposed to having it auto-adjust based on the
surrounding noise. This will prevent amplifying background-noise when
nobody is talking. Also it is desireable to record a couple seconds of
silence before starting the recording, so you can use audacity or some
other audio-processing tools to filter out the background noise in
postprocessing. But whether you need to filter the audio  of course
depends on fhe quality of the microphone and also on the surroundings.
But consider using audio-level compression (to reduce difference
between softly spoken and loudly spoken parts). If you had to cough
right into your microphone, make sure to edit that out!! There are so
many videos on yt where suddenly the person coughs right into the
microphone, maxing out the input level resulting in a much much louder
sound than the regular speech. Avoid that. It is easy to edit audio so
there is no excuse for leaving that in. Not saying that you must not
cough or sneeze or do whatever in your videos, but when that results
in a vastly different volume level, then you better do some
postprocessing.

All below is very special, so it's ok if you stop reading here :-) I
thus put my greeting here..

ciao
Christian

What  I use doesn't necessarily fit other users, esp. Windows users
will very likely use different tools, but nevertheless software I
used:
* Aegisub/a subtitle program - to add any on-screen text and static fades
* pitivi (for easy cross-fades or simple effects, also make sure to
render losslessly in intermediate steps) - but that is a little buggy)
* avidemux2 (to cut out the black frames pitivi adds after a video
ends on the timeline, despite the rest continuing, and also to play
with filter settings while having a preview)
* mplayer(2) - to do cropping & resizing, denoise (hqdn3d) &
color-filter (eq2) if necessary, and combine with the subtitle file.
pipes yuv4mpeg format to
* x264 for encoding. For screencasts, use lossless mode, that
compresses really well. Otherwise make choice between your upload
speed (how long it will take to upload your video to youtube) and
encoding speed (how long does it take to encode the video). I
typically use --crf 20 and preset --veryslow - as I have rather slow
upload, and that version can then also be used to offer a
direct-download of the video.
* audacity for audio editing (if necessary)
* mkvmergegui (mmg) to mux the video and the audio together, also to
sometimes time-scale the video (i.e. adjust framerate) to match the
length of the audio. [2]

[1] I personally use ffmpeg in combination with x264 - ffmpeg just is
the swiss-army-knife and allows to capture anything from either parts
of the screen or in whole, and x264 just offers a great performance
and small files in lossless mode. use a preset that will not tax your
CPU while recording and will give you a comfortable filesize (only
matters for longer screencasts). i.e. start with --fast, and if it
uses too much cpu for your taste, use --veryfast or --ultrafast
instead. If you have plenty of CPU to spare, use --normal or --slower.
But use --qp 0 for lossless format.

[2] also did use a timestamp files to increase framerate on parts of
the build-demo recording to create a fast-forward effect (framerates
up to  900fps) - but of course only as an intermediate step - that
accelerated version was encoded with standard framerate (the additonal
ones were simply dropped) - as I was using lossless format, no loss in
quality \o/)

Example commandlines:

* play video with mplayer and apply filter & a subtitle file
(auto-loaded if it is available with the same basename as the video
file) and encode with x264

mplayer -nosound -benchmark -vf hqdn3d -vo yuv4mpeg:file=>(nice -n 18
x264 --demuxer y4m --crf 20 --preset veryslow --tune film --output
~/tmp/encoded_version.mkv - 2>~/tmp/x264.log )  inputfile.mp4

(denoising filter hdn3d with defaults settings applied, and encoded in
lossy mode (--crf 20))

* screencapture, lossless of part of screen - I prefer to call x264
manually than to use ffmpeg's builtin switches
ffmpeg -loglevel warning -f x11grab -r 15 -s 1202x698 -i :0.0+0,102
-vf pad=1280:720:39:11 -f yuv4mpegpipe -pix_fmt yuv420p - | x264
--demuxer y4m --preset veryfast --qp 0 -o buildsession.mkv -

I recorded a window of 1202x698 pixel, starting at the very left, 102
down of my screen (I placed my terminal so that window borders are out
of screen to the left & bottom), and then pad it to 1280x720,
centering the captured are. Record with a framerate of 15 frames per
second. ffmpeg hands it over to x264 in yuv4mpeg format in the yuv420p
colorspace.
x264 encodes it with lossless setting, and the veryfast preset (don't
bother to create a small file, just work quickly).

-- 
Unsubscribe instructions: E-mail to marketing+help@global.libreoffice.org
Problems? http://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: http://wiki.documentfoundation.org/Netiquette
List archive: http://listarchives.libreoffice.org/global/marketing/
All messages sent to this list will be publicly archived and cannot be deleted