'image2' not 'image 2'
>But I can give those manual commands on the single frames gotten with
>$ ffmpeg -i VIDEO.mkv -f image 2 DIR/SINGLE-FRAMES-%5d.png
>(that ffmpeg command gives you exact frames, 30 per second in my case,
>in PNG format)

The Cinelerra Motion Stabilization PART 2 video, that I am preparing
will, hopefully, be clear real world example illustrating, or at
least giving a hint for the scenario that Cinelerra devs, I guess,
will sooner or later embark on, by the necessity of progress in this matter
(Pls. don't attribute it to my wish when this, I bet, verifies, it is
purely logical that this be the next step for our good
ole devs who make this great program!).

What I did (and I appologize for more typos in the first successful
message on this "Motion plugin great, but perspective correction
impossible" topic, such as 'image 2' instead of 'image2' in one of the
ffmpeg lines), is, now from my command line history I am taking, the

ln -s ../SOMEWHERE/on_my_box/complete_half_our_video.mkv \
mkdir   Try04/  
ffmpeg -ss 123 -i KrugTrg_Cinelerra_IM.mkv -f image2 -frames:v 300 \
gwenview Try04/ &
ffmpeg -r 30 -i  Try04/kt-%5d.png -map 0:v -r 30 -c:v libx264 -g 10 \
mkdir Try04_RE/
ffmpeg -ss 0 -i KrugTrg_Cinelerra_IM_SAMPLE.mkv -f image2 \
gwenview Try04_RE/ &

I just link the half our video and called it, by virtue of the
symlink, KrugTrg_Cinelerra_IM.mkv, in the current dir (IM for

There's three ffmpeg lines above. Devs, pls. skip this, this
explanation is for the newbies.

The '-ss 123' in the first ffmpeg line starts churning out images from
2min 3 sec of the video, and '-frames:v 300', since video is 30fps,
limit the output to (300/30 is) 10 seconds of video. That is our sample.

gwenview is the fine GNU/Linux program I view those images in (I
really love GNU/Linux, and the licence and the freedom of it is
sacred, and under attack, if you didn't know, but I could only try and
make a new topic on this, cannot talk of that in this topic).

The options '-r 30' in the second ffmpeg line is given two times, and
I don't know it both are needed. What I do know is, it's pretty
perfect ffmpeg program at work here, and I notice no losses really
when turning videos into pics and then pics into videos, even a few
times!  '-g 10' is non-essential in this story, for completeness, it
makes my videos cut-able to around 0.1 seconds intervals, usually. Can
omit it here.

Because, what I got with the third ffmpeg line, where, sure, the '-ss
0' can be omitted, just like no '-frames:v number' option wasn't
needed, because we already had exactly the 300 frames to make, is
exactly the SAMPLE video that is drawn out of the exact 300 frames
from the video, regardless, or at least very nearly regardless
(i.e. without perceptible losses), of the number of '-f image2'
conversions to pics and than back to video.

That is why I will post the video to get users' and devs' opinion in
this matter, if possible.

I'll do so because I believe that my steps can be fine replicated by
users' and devs' by simply downloading this video to their
computers.  ...And doing so much more than what I was able to, the
devs, if they decide to wrestle a little with the Motion plugin... or
at least, regular users, reasonably well understanding my concerns.

I'm sure that adding the perspective correction (or differently named
feature, but doing that) to motion stabilization is not itching only

Before I upload the video, I need to explain here that I could
possibly be at the mercy of Google/Youtube local bullies here. No more
words on them, it's too ugly note in a fine story.

(I do have to notice that uploaded 1440x1080 PART 1 video wasn't fully
shown as such, but only as 960x720, when I downloaded it, but I have
to say that the losses are not much perceptible in the part that I
will point you in the next paragraph. So possibly not really a problem
_here_ but they did downsize with serious losses my videos in the
past! Oh yes!)

Just, I will upload my flowstamped video (flowstamp being my
bash-script program, that I intend to release soon, flowstamped very
marginally, too few even and smooth -but not glossy!- surfaces to
notice parts of words www.CroatiaFidelis.hr and Miroslav Rovis along
with the sha256 number that identifies the original Samsung Galaxy II
camera video of 0:27:53 length, the intention on my part is not to
bother viewers with too visible stamp) as was taken and as it is
preserved. No losses. 1920x1080, true high resolution video, as
anybody can see in the Cinelerra Motion Stabilization PART 1, go to
5:40 of the video, the "Boos and Three Claps for the President"
part. Pls. do and through comparison of the small area covered, and
the loss, here, which is certainly significant loss, with the screen
capture command...

[[ This is that command:
ffmpeg -f pulse -i default -f x11grab -s xga -r 25 -i :0.0 \
 -vcodec libx264 -preset ultrafast  -threads 0 \
 Screen_`date +%y%m%d_%H%M`_`hostname`.avi

...you should figure out that the video of the 10 seconds that I am
about to post should come out pretty neat for you.

Thinking more about it, I can't really tell if their reediting the
video, automatic (I have also witnessed purpuseful real time, as my
video comes, usually malicious, reediting by those non-gentlemen)
usually, would change the ordering of frames.

But let me upload now that video and I'll continue from this dilemma
just expressed.

I have to say just one thing. I am never certain when they could go
bonkers and cut my very internet connection completely down and out
for all intents and puposes (they did so on some two occasions), so in
case you don't read from me more, that (or worse) could be the matter.

Miroslav Rovis
Vankina 4, Zagreb,
m.ro...@inet.hr #rarely used, regime-pestered
miro.ro...@gmail.com #rarely used, regime-pestered
+385 (0)1 660 2633
+385 (0)91 266 0202
#=== was sentenced to jail, 2 ys on parole, for political views, essentially 
# but no English there at this time

Cinelerra mailing list

Reply via email to