Today I attended the Open Subtitles Design Summit 2010, and just wanted to quickly jot down some of what i took out of it.
The first discussion was on how to connect the subtitle with the video, so if someone has a video, they can search for the subtitle info without having to write it themselves. A few problems with this:
– Copyright issues. If someone makes a video, someone else cold subtitle it, put it online, without the content owner even knowing that this happened.
– Synching multiple video links to the same video with a key, a unique key, so there doesn’t exists two subtitle files for the same video. The problems with this is if a video is essentially the same by name, but one has commercials, thus is no longer the same from a technical standpoint. An idea to solve this is to get a finger print of the audio data, and as far as subtitles are concerned, that is the exact same video.
– The final main problem I took from this is where to host this subtitle data that people could look up. Do you do it in one place like a database, or do you spread it around like a torrent.
Second discussion was on video meta data. That term can be thrown around without a clear definition.
– The main thing I got from this is that there are three types of meta data. Data about the video, like the size, length, or other technical details. Data about the video’s contents, like the people in the video, the people that made it… content. The final one, which is timed meta data, like a subtitle starting at time x and ending at time y. The concept of timed meta data helped me understand it, because it made me see the object inheritance of meta data. If you look at it this way, the video is data, and the subtitle is meta data, and the start and end time is meta data about the subtitle, and not the video itself. So maybe there is only two types, with subtitles being about content, and the start and end being technical meta data about that subtitle.
– Another concept that interested me is that some of this data will never change. Like director. The director of a film will always be the director, but something like rating will fluctuate. Both are not new types of meta data as they are both data about the content, so content or technical meta data can be fluid OR concrete.
Fourth discussion was about popcorn.js and where we can go with it.
– We need to consider our xml format, and make it standard, because we met pad.ma which is doing something very very similar, in a very similar way, so we have to take a closer look at the xml and see how we can create a standard between at least the two of us, to get the ball rolling.
– Also, because of the discussion on meta data, I have a clear description of what popcorn is… timed meta data.
– Having a command to link into the api of what people are listening too. Just a new command to add to the library.
– Something else that is really interesting is a working standard, which is not yet implemented, but is a timed event listener called oncuechange… Basically, do this function at this time. It can be found here which is very exciting and could improve the foundation of popcorn.js.
Fifth discussion was on openindie service. Which is like a hub for independent video makers to get their videos out there, and send screeners of their work to people.
– They were looking into getting subtitle data much the same way popcorn.js is from universal subtitles, and I think it is important for popcorn and openindie to work together with universal subtitles to get the same data, so it only has one format. This would probably be different than the subtitle info used by pad.ma and popcorn. The difference being xml and json, and different standards for each when it comes to subtitles and semantic data.
Final discussion was the last little bit on what can be done in the next 3 months, and what can be done in the next 12 months.
– the best thing from this was an idea where we oculd use the audio api to get the raw sound data of a video, and read it, then play one sentence, then pause, so whoever is entering the subtitle can keep up with the video. Write one sentence, hit enter, write the next, etc. This would be ideal and very intuitive. Talked to David Humphrey about this, who is the lead developer of the audio build, who recently had a discussion on this, and he gave the idea of writing a library that would control and deal with when sound of a video changes in frequency, and then you can fire events, or do something else with it.