transcripts/subtitles #56
Comments
|
debconf has something for this too, and it's organized entirely outside of the whole record/review workflow: https://wiki.debconf.org/wiki/Videoteam/Subtitles I can probably contact the guy who set that up for debconf and ask if he's interested in doing it for FOSDEM, too. If so, he might be able to do that for 2017 still. |
|
Check out http://amara.org/en/community . Source available at https://github.com/pculture/unisubs . |
|
Emanuil will recruit remote volunteers for this. |
|
Here's how Include an srt subtitle stream in the mp4. Requires a player capable of playing the subtitles. ffmpeg -i infile.mp4 -i infile.srt -c copy -c:s mov_text outfile.mp4 |
|
Or burn it into the video: Use the libass library (make sure your ffmpeg install has the library in the configuration --enable-libass). First convert the subtitles to .ass format: ffmpeg -i subtitles.srt subtitles.ass Then add them using a video filter (via libass): ffmpeg -i subtitles.srt subtitles.ass ffmpeg -i mymovie.mp4 -vf ass=subtitles.ass mysubtitledmovie.mp4 |
|
https://github.com/FOSDEM/video/wiki/Subtitles-and-mp4-metadata collects the above info and more in a somewhat structured way. |
|
I should add that I have vague plans of adding support for burning subtitles into videos into SReview, too. Downside of doing that would be that the SReview infra would need to be kept alive for as long as subtitles are being written... |
|
I have investigated subtitles rather thoroughly. We definitely want something else than burnt in subtitles (at least for FOSDEM, but I guess also for other potential uses). The state of the art on subtitles is:
What we need from sreview in order to make subtitles succeed imho is a few things:
|
|
On Sat, Dec 30, 2017 at 10:33:57AM +0000, Mark Van den Borre wrote:
I have investigated subtitles rather thoroughly.
Me too, actually. One result of that was
https://grep.be/blog/en/computer/play/Adding_subtitles_with_FFmpeg/
We definitely want something else than burnt in subtitles (at least
for FOSDEM, but I guess also for other potential uses).
Yes. Burnt in subtitles means you have only one, and it can't be
switched off. With VTT subtitles, you can have multiple ones, and the
user gets to decide whether to switch them on (and if so, which one).
At least VLC, mpv, and mplayer support that.
The state of the art on subtitles is:
• separate webvtt subtitle files for web based use (html5 standard, fairly
well supported on all browser platforms)
WebM containers can incorporate them, too, which I think is preferable
from a usability point of view (you only need to download one file), but
not so from an infrastructure point of view (you need to rewrite the
output file over and over again, every time someone writes new
subtitles).
• embedded tx3g subtitles for offline playback (fairly well supported across
most decenf offline players)
Not sure if tx3g is the best option, but then for mp4 files it might be
(haven't investigated that bit in too much detail, tbh).
What we need from sreview in order to make subtitles succeed imho is a few
things:
• embed penta talk metadata in the mp4 metadata. See https://github.com/
FOSDEM/video/wiki/Subtitles-and-mp4-metadata for a trivial way to do that.
We'd need at least the penta id, but more would be useful. I'll come up
with more details after a closer look at amara, and I'll file an sreview
issue. You might already want to start playing with this though, because
the advantages of metadata embedding go far beyond just subtitling!
That's already done :-)
See
https://github.com/yoe/sreview/blob/master/lib/SReview/Video.pm#L576-L580
for more details. Yes, SReview sets those metadata items (if it has any;
for FOSDEM, the schedule import script grabs them from penta)
• expose ready-to-subtitle mp4 urls (json url+ timestamp?). Ready-to-subtitle
in our case means they have their definitive spot on video.fosdem.org.
Video.fosdem.org needs to stay simple, clean content delivery only.
Yup, agreed. The URL scheme will be very similar to last year's.
• (probably) submit the finished video urls directly to amara.org.
That's certainly something I can look into.
…--
Could you people please use IRC like normal people?!?
-- Amaya Rodrigo Sastre, trying to quiet down the buzz in the DebConf 2008
Hacklab
|
|
> • separate webvtt subtitle files for web based use (html5 standard,
fairly
> well supported on all browser platforms)
WebM containers can incorporate them, too, which I think is preferable
from a usability point of view (you only need to download one file), but
not so from an infrastructure point of view (you need to rewrite the
output file over and over again, every time someone writes new
subtitles).
I think we are on the same line here. Embedding it into the mp4 or webm
file is no use for browser use, but very useful for offline use.
• embedded tx3g subtitles for offline playback (fairly well supported
across
> most decenf offline players)
Not sure if tx3g is the best option, but then for mp4 files it might be
(haven't investigated that bit in too much detail, tbh).
It's the only widely supported option for offline mp4 embedded subtitles.
Webvtt is better to embed in webm. Ffmpeg will auto convert mp4 wtih tx3g
subtitles embedded to webm with webvtt subtitles embedded.
> • embed penta talk metadata in the mp4 metadata.
That's already done :-)
Great! That's really good news! Can I have an example mp4 file with penta
metadata embedded in the way sreview currently does it please?
> • expose ready-to-subtitle mp4 urls (json url+ timestamp?).
Ready-to-subtitle
> in our case means they have their definitive spot on video.fosdem.org.
> Video.fosdem.org needs to stay simple, clean content delivery only.
Yup, agreed. The URL scheme will be very similar to last year's.
> • (probably) submit the finished video urls directly to amara.org.
That's certainly something I can look into.
The json is probably the most useful bit to get ready in the short term.
Tbh, for the limited number of keynotes & main tracks we plan to do, worst
case we do the amara submission by hand.
It _would_ be important wrt to focus review on Janson and K1.105 Lafontaine
mp4's first before anything else... Don't want to block the review queue
with less important talks that don't get subtitled. I guess that is already
how sreview works though?
Regards,
Mark
--
Mark Van den Borre
Hogestraat 16
3000 Leuven, België
+32 486 961726
|
|
On Sat, Dec 30, 2017 at 03:15:25AM -0800, Mark Van den Borre wrote:
> > • separate webvtt subtitle files for web based use (html5 standard,
> fairly
> > well supported on all browser platforms)
>
> WebM containers can incorporate them, too, which I think is preferable
> from a usability point of view (you only need to download one file), but
> not so from an infrastructure point of view (you need to rewrite the
> output file over and over again, every time someone writes new
> subtitles).
>
I think we are on the same line here. Embedding it into the mp4 or webm
file is no use for browser use, but very useful for offline use.
Actually, browsers *can* play embedded VTT subtitles from WebM files,
but it may require some javascript help.
> • embedded tx3g subtitles for offline playback (fairly well supported
> across
> > most decenf offline players)
>
> Not sure if tx3g is the best option, but then for mp4 files it might be
> (haven't investigated that bit in too much detail, tbh).
>
It's the only widely supported option for offline mp4 embedded subtitles.
Right.
Webvtt is better to embed in webm. Ffmpeg will auto convert mp4 wtih tx3g
subtitles embedded to webm with webvtt subtitles embedded.
Sure, it converts loads of stuff :-)
> > • embed penta talk metadata in the mp4 metadata.
> That's already done :-)
>
Great! That's really good news! Can I have an example mp4 file with penta
metadata embedded in the way sreview currently does it please?
Don't have one on hand right now, and don't have too much time to look
into generating one. If you really need one, I can look into generating
one later.
The code that sets the actual metadata is in
<https://github.com/yoe/sreview/blob/master/scripts/transcode/sreview-transcode#L294-L298>:
we currently set the talk's title, event (name; e.g., "FOSDEM 2018"),
speakers (names; format like "Wouter Verhelst, Richard Hartmann and Mark
Van den Borre"), track (name; what's entered in penta for track name),
and date of the talk.
It should be possible to add the penta ID and other things, if needs be
(just needs to be pulled from the db and added to the foreach loop
there).
> > • expose ready-to-subtitle mp4 urls (json url+ timestamp?).
> Ready-to-subtitle
> > in our case means they have their definitive spot on video.fosdem.org.
> > Video.fosdem.org needs to stay simple, clean content delivery only.
>
> Yup, agreed. The URL scheme will be very similar to last year's.
>
> > • (probably) submit the finished video urls directly to amara.org.
>
> That's certainly something I can look into.
>
The json is probably the most useful bit to get ready in the short term.
Tbh, for the limited number of keynotes & main tracks we plan to do, worst
case we do the amara submission by hand.
Right, although I think it would be a disservice to our devrooms if we
allow transcript/subtitling for main track talks but not for them.
It _would_ be important wrt to focus review on Janson and K1.105 Lafontaine
mp4's first before anything else...
This sentence misses a word, and I can't parse it.
Don't want to block the review queue with less important talks that
don't get subtitled. I guess that is already how sreview works though?
In light of the above, I'm not sure what you're asking here. Please
clarify.
Regards,
Mark
--
Nitpick: it's dash-dash-space-newline, not dash-dash-newline. My mail
client (and github) missed that your mail ended here ;-)
…--
Could you people please use IRC like normal people?!?
-- Amaya Rodrigo Sastre, trying to quiet down the buzz in the DebConf 2008
Hacklab
|
Can we have an input field that allows for uploading VTT or SRT files? You would (in the first iteration) defer the creation of subtitles from the playback. The JavaScript snippet used to show subtitles is already in the wiki. |
|
Hi André,
On Thu, Feb 10, 2022 at 11:10:59AM -0800, André Jaenisch wrote:
https://github.com/FOSDEM/video/wiki/Subtitles-and-mp4-metadata collects
the above info and more in a somewhat structured way.
Can we have an input field that allows for uploading VTT or SRT files?
(I wrote mine manually).
There is currently no functionality in SReview (the system we use for
managing video) to handle subtitles, at all. It's on my low-priority
TODO list to handle this somehow; if you are interested in writing the
necessary patch, that would be most welcome! the code for SReview is at
<https://github.com/yoe/SReview> and at
<https://salsa.debian.org/debconf-video-team/sreview>.
In my ideal world, subtitles would be stored as files on the SReview
system somewhere, and there would be an extra state where the subtitles
would be muxed into the output files as a proper subtitle stream with
the correct language tag set, etc (if the container format supports it).
In the mean time, I think the ideal way forward is for you to upload the
subtitles into pentabarf as an attachment; you go to the "Files" tab,
click on the "Add attachment" button, and select "Other" as the
"attachment type". They will then be copied to the website.
…--
***@***.***{be,co.za}
wouter@{grep.be,fosdem.org,debian.org}
|
|
Hey there
Oh, okay. At least it is on your mind.
I took a look at the code, but don't speak Perl. Sorry.
Doing that already. |
|
I have done automatic subtitles with whisper here: |
For a future edition...
Let's see if we can encourage people to easily build transcripts/subtitles.
See https://wiki.c3subtitles.de for inspiration.
The text was updated successfully, but these errors were encountered: