[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [Debconf-team] [Debconf-video] subtitling 101

In-line :-

On 02/05/2016, alberto fuentes <pajaro@gmail.com> wrote:
> The wikipage is more or less up to date with current practices :)
> https://wiki.debconf.org/wiki/Videoteam/Subtitles
> I dont think there is anything technical preventing more people
> contributing with the subtitles. In the past contributors were expected to
> know git and have an alioth account, but last year we moved (mostly) to
> web, removing any kind of hindrance new contributors might have or
> technical knowledge they need to start
> As I see it, the main big problem is that people that is able to make great
> subtitles, usually dont need them. The fact that these users dont overlap
> with the people that see great need in subtitles reduces the volunteer pool
> greatly
> The second big problem is that making subs is a lot of hard work that not
> many people realize. I estimated that it takes from 1h to 2h for every
> 10min of finished subtitles. And this takes into account that you are
> proficient while doing it. It can be even slower if you are a slow typist.
> People starts with a lot of energy, but get discouraged soon when they
> realize the pace to finish a single subtitle for a single video

Dear alberto,

Please do not take my queries as anything other than queries. I think
the subtitle team does highly valuable work. I know about amara but
have found using subtitleeditor, aegissub or even the basic
text-editor (leafpad) better than amara as amara needs constant live

I have made partial subtitles (not debconf) but other videos but the
biggest issue for me has been time-stamps and frequency. From the
script shared from unix.stackexchange.com, it seems this task could be
automated so for a potential off-line subber, he could have an empty
file with time-stamps and a base frequency for the subtitle (say 5-6-7
seconds) it would be more a matter of simply filling in. Does anybody
know of a script which will take the inputs of duration of video and
frequency and output an .srt file ?

Another query I have, either it's subtitleeditor's fault or .srt
format (or something I don't know) but if I change the time on one
subtitle, the other lines do not change.

00:00:00,000 --> 00:00:01,488
... wanted to be able to use

00:00:01,488 --> 00:00:03,284
Thunderbird and GnuPG together with Tor,

Using the sample as before

Now if I change the time-stamp of the first line :-

00:00:00,000 --> 00:00:02,000
... wanted to be able to use

00:00:01,488 --> 00:00:03,284
Thunderbird and GnuPG together with Tor,

Now in this modified version, the time-stamp of the first line changed
from what was original and if I were to play it, the first line of the
subtitle would merge with the second line when it crosses 00:00:01,488
but there doesn't seem to be a way to fix that automatically. If
somebody knows a way to automatically check and either fix or show the
user where such merging, overlapping happens it would make it also
easier. This is more for fixing subs. though, not creation so much.

Most of my time in subtitling I have spent in projects have been
getting time-stamps fixed aka syncing.

The most I have done of original subtitling though is of 20-25 minutes
and that too over 2-3 nights as rightly pointed out, it takes lot of
time to make subtitle for even 10 minutes of video, that is true as it
takes multiple tries to make sure that the video and subtitle sync
right as well as to get to know what a person meant to say, at times
the inflection of the voice goes down and you have to guess what word
the person said which is dangerous as your mind starts to conjecture a
word where it could be something else s/he said.

> Another problem is that debconf subtitles 'hotness' is highly volatile.
> Workshops and tutorials are maybe the exception, but talks value decreases
> over time, aka, less people is interested on projects updates one or two
> month down the line

Probably, but it probably is influenced by also who is speaking in part as well.

> And the final problem, looking at the numbers in the past, the volunteers
> and casuals people interested to help peek during the actual debconf and
> the week after (maybe when lwn publish the link the talks). This was
> slightly fixed last year by the hard work of people form the video-team by
> having the video available on the same day they happened... :)

That is part of the solution. I think if it is possible to  have the
video on the same day, then it is also possible for volunteers to do
subtitling, but this would probably also go hand-in-hand with what
people liked/enjoyed during the talk. Sometimes the Q&A is where
subtitles are needed the most and they are not available. I am not
blaming the subtitle team, it's either something they are not aware
that people need them also there or people thinking just the
essentials (the talk)  are enough when in reality, it's sometimes the
Q&A which are more revelatory than the talk itself.

> ...but even long-term volunteers had to divide their attention between
> helping with subtitles and watching that day's talks live or meeting new
> people or eating cheese...

True, true know what you mean, downtime is also needed as well as
keeping oneself freshened for the next day as well.

> We are open to new ideas to improve the way we work or new hands or both :)

I don't have any concrete proposals or ideas other than the queries
and am open to spending sometime on the subtitle team as well provided
we get the video on the same day (the videos would probably have to be
down-sampled so they are easily fit in user's machines as the video
team probably shoots the videos in raw-video and then puts them in
720p or 1080p .webm which people can use them.

I will be seeing the aegis sub youtube video that you have linked to
in the wiki article.

> Well see how it goes this year :o)
> Cheers!


This is not a mistake. The video recording starts half-way through a
sentence, so it's an accurate representation of what is in the video.


Dominic, I was just sharing it as an example, it isn't and wasn't mean
to be accusatory in any way and if you feel I was being in anyway,
then let me apologize as that wasn't what I was pointing at. My main
was to point at the time-stamps and the ordering that a potential
subber has to do/contend with.

Look forward to learning from all :)

          Shirish Agarwal  शिरीष अग्रवाल
  My quotes in this email licensed under CC 3.0
EB80 462B 08E1 A0DE A73A  2C2F 9F3D C7A4 E1C4 D2D8

Reply to: