2008-01-08, 21:56 | Link #1 |
Aegisub dev
IT Support
Join Date: Dec 2004
Location: Florianópolis, Brazil, Pale Blue Dot
Age: 38
|
Parallel colaborative fansubbing proposal
Today, I had an idea that some people in #aegisub judged to be interesting. After a brief discussion there, I've decided to post it here to gather the opinion of more fansubbers.
I already expect a lot of flaming, especially from the "speedsubbing sucks" groups, although this proposal would benefit everyone, not just speedsubbers. The idea is to have some sort of client-server system (with the client integrated into Aegisub) so multiple people can work on different stages of the fansubbing pipeline of a single episode all at the same time. Basically, there's a server that you connect to (with login and password). The server lists all active projects. You pick one, and assign yourself a specific position, for example, "timer". Then, every connected member can start working on it as work becomes available. It all starts with the translator, of course. As he watches, he will enter lines in a text editor-like interface (with a few additions, such as F1-F12 keys to quickly switch current actor). As soon as a line is entered, it starts to "count down" (say, a minute). The translator can still change the line while the countdown is going. When the countdown is over, the line is sent to the server as "ready for the next stage". That way, the translator doesn't have any extra work, and the next stage can start as soon as the translator has typed in a few lines. Alternatively, there can be a "flush" button, to flush the current pending lines. This is up for debate. After the lines were submitted, the server understands them as "translation OK, waiting timing and editing". On that moment, both the editor and timer will have the line available for them. The timer will only be able to change the actual times, while the editor only the actual text, therefore their work is completely independent. As they are done with lines, they are also auto-set for the next stage, or flushed manually. Maybe typesetting could also be done in parallel at this stage. After that, if typesetting was already done, all that is left is translation checking and QC. Translation checking can happen immediately after editing, and might be optional (server-side option). The final stage, QCing, will likely be slightly different: the program will wait for a reasonable ammount of lines to be "ready", and, once they are, tell the user that it can has more stuff to watch. Then the user watches the fragment and hits a key whenever he sees a problem. The program then pauses the video and asks what kind of problem it is, translation, editing, typesetting, etc, and the user picks it, with optional comment. The current timestamp is marked as "in need of revision", and the appropriate person assigned to the episode is notified of such. If you've been keeping up with my logic, it should be clear that such a system would allow an entire episode to go through the entire fansubbing process, including more than one QC, in an hour or so, if the team is all available and works at typical speeds. Also, the server would work in a somewhat SVN-like fashion, preventing version conflicts and keeping a history of all changes. It's up to debate how would "late revisions" work... For example, if a translator wants to change something after it's already been edited. Either way, it shouldn't be too hard. Now, this is just mindstorming. I had the idea and decided to share with the community. This all could feasibly be implemented in Aegisub after the planned 2.10 reformation (yes, we still need to release 2.00 before, we KNOW ), but we will only really consider it if there is sufficient community support for it. The server itself would be a lightweight cross-platform program, which would be administered remotely by clients with admin privileges. A single server could potentially host many different fansubs, and each fansub could be organized in projects, and each project in episodes. Each user would them assign themselves in a position for some project (or maybe the admin would?) in order to work on it. There's also some question on how to share the raw - should that be done manually, or should the program deal with that in some way? That way, once the "distro" client has the raw, it could stream it to other users, so they can start working as they receive it... Although that's probably overkill. Anyway, I'm not promising anything! This is just an idea, for now. It MIGHT become reality someday, who knows? But how the community reacts will largely influence that. Thanks for your time and sorry for the long post! [EDIT] BTW, the mentioned 2.10 reformation is mostly source-wise, not interface-wise, but will allow things such as multiple subtitle files loaded at once, which is pretty much mandatory for proper Mac support. [EDIT2] Here's the log from #aegisub: Spoiler for Long IRC logs:
Last edited by ArchMageZeratuL; 2008-01-08 at 22:18. Reason: Clarifying the 2.10 reformation. |
2008-01-08, 22:21 | Link #2 |
Junior Member
Join Date: Jan 2008
|
Like I mentioned on IRC, basing this on obby (the collaborative editing part of gobby) would give you about 90% of what you're describing here. (For those of you who weren't on IRC: In Flomp, we use gobby for a good bit of the subbing process, and it works really really well. In actual practice, using a system like this, we can finish an episode in 3-5 hours.)
A few suggestions: * Instead of the minute wait for the lines to be sent to the server, it'd probably be easier (and more useful) if the text was sent as it was being typed. Also prevents race conditions when two people are editing the same line. * Having the server be aware of what stage a line is at is a really good idea - that should be developed further, it'd be really useful, especially if you allow the user to re-label lines that need to be redone in some earlier stage. * gobby already has a lightweight, cross-platform server available - one more reason to use it |
2008-01-08, 22:58 | Link #3 |
Senior Member
|
Great idea, impossible in implementation.
p-static... What group are you with? I'm a very experienced fansubber and I've never heard of a 3-5 hour turnaround unless you mean an uber speedsub with no editing or QC. The fastest I've ever done is 10 hours, and believe me, that's considered extremely fast. -Tofu |
2008-01-08, 23:07 | Link #4 |
Aegisub dev
IT Support
Join Date: Dec 2004
Location: Florianópolis, Brazil, Pale Blue Dot
Age: 38
|
Impossible?
I don't even see anything as particulary CHALLENGING... It's just a LOT of work, but none of it is *hard*. Or do you mean something in particular that I've overlooked? |
2008-01-08, 23:14 | Link #5 |
Junior Member
Join Date: Jan 2008
|
Yeah, it's a speedsub, but the quality is actually pretty decent. We have a fairly efficient pipeline set up, where everybody can work simultaneously, which takes off a lot of time. Releasing softsubs also helps, since we can encode in parallel with everybody working. That breaks down somewhat if we need to do any complicated typesetting that can't be reliably done in a softsub, of course, but for the most part that isn't a problem.
In the end, the speed of the process is basically TL time + timing time + edit / tl check / qc / typesetting time, since everything after timing can be done simultaneously. So with collaborative editing built into aegisub, that could be improved even more. :3 |
2008-01-09, 09:54 | Link #7 |
aka Raw Provider
Join Date: Aug 2004
Age: 37
|
I agree with Tofusensei.
Although, go ahead and try, if you manage to get it to work, I am more than willing to try it. Shame that every one lives in different places of the world, which slows the things downs no matter what tools you are using. |
2008-01-09, 11:12 | Link #9 |
Senior Member
Join Date: Dec 2007
|
If you're willing to take the time to implement this, then go ahead. It sounds like it could, in most group, at least speed up the process somewhat.
The only problem for some groups might be a place to put this server on, but I guess they all have at least one member who stays online 24/7. In any case, I don't like the countdown parts. |
2008-01-09, 11:53 | Link #10 |
My E-Penis > Your E-Penis
Fansubber
Join Date: Feb 2006
Location: Brussels, Belgium
Age: 38
|
i guess it would need some kinda of color code to show which lines are final and which not. but even then, i personally don't work from beginning to end, in a linear way. as i edit, i skip around the script, change stuff everywhere, etc. i guess it might work for like, translating and timing at the same time or so, but that would mean that the timer would have to time without having tled lines, something not many timers like to do. it would also have to mean that the people are able to work at the same time, and that they can communicate directly as to not interfere or change each other's work.
in short, it's an interesting idea, but not worth pursuing. i don't think anyone but speedsubbers would successfully make use of this idea. it's too risky for groups that prioritize quality over speed.
__________________
|
2008-01-09, 12:02 | Link #11 | ||||
Aegisub dev
IT Support
Join Date: Dec 2004
Location: Florianópolis, Brazil, Pale Blue Dot
Age: 38
|
Quote:
Quote:
The whole POINT of this system is that even if you NEED a particular task before yours, you don't need it to be DONE: you can work on the partial work. Quote:
Of course, people being able to work at the same time is the biggest issue. Quote:
But, yes, speedsubbers would benefit from it the most. Now that I think about it, I can imagine that non-speedsub groups would want to oppose this idea on grounds that they don't WANT speedsubs to release so fast... Thoughts? |
||||
2008-01-09, 12:09 | Link #12 |
Ana-chan~
Join Date: May 2006
Location: Netherlands
|
'quality' groups get a lot of their work done out of aegisub. So it's probably not gonna make a lot of difference for those groups, as aegisub isnt really what's the bottleneck for them. Though.. I like more efficient/parallel QCing (in case of more QC'ers that is.. 1 good qc'er should be fine too :P), and the 'tracking changes' stuff.
|
2008-01-09, 12:18 | Link #13 |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Frankly, I would think this would be best implemented in a hosted website application, ala google aps, instead of a stand alone program.
Implement the whole thing in java or perl or some web-application construction toolkit, host it with some space on a cheapy webserver and THEN you'll have the easy for anyone to use, always on, anyone can fansub simultaneously solution that might be interesting to me. Then anyone with a web browser, from anywhere, could work on the show, even at the same time as you propose. P.S. While you're at it in Aegissub, implement a chat function for basic chatting with people currently working on the same server. Or maybe even a meta-server which links all the smaller ones, thus creating a basic IRC style network with each channel being a working group (groups could talk to one another for help/advice, etc).
__________________
|
2008-01-09, 12:25 | Link #14 | ||
Ana-chan~
Join Date: May 2006
Location: Netherlands
|
Quote:
Quote:
|
||
2008-01-09, 12:32 | Link #15 |
Aegisub dev
IT Support
Join Date: Dec 2004
Location: Florianópolis, Brazil, Pale Blue Dot
Age: 38
|
I can't see why would a web-app be any better.
It would be considerably harder to code, wouldn't have access to all of Aegisub's features... and how would you typeset to video and time to audio with that? As for the chat, a simple chat is trivial to implement, so it'll probably be there. |
2008-01-09, 12:42 | Link #16 |
Junior Member
Join Date: Jan 2008
|
That brings up another thing I had forgotten about: another *really* nice feature in gobby that should definitely make it into aegisub if this happens is a built in chat thingy.
Also, I don't really see the need to have strict restrictions on what anybody can do to the script at a given time. If you don't want somebody messing with your stuff before you're done with it, there's nothing stopping you from just asking them to hold off for a bit. ^_^ |
2008-01-09, 13:06 | Link #17 | |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Quote:
Now your timer no longer can give the excuse "But I'm on vacation at my grandmother's house for 3 weeks!" As long as grandma has internet explorer and DSL, they no longer have an excuse . As for the chat suggestion, the point of having a built in chat-room is to have the entire fansub development platform in one program. Timing/translation/edit/qc, and a crucial part of that is communication between the people involved. No need to reinvent the wheel of course, it might be simplest to just include a basic IRC server/client protocol in Aegissub itself. Or you could go more modern and implement it with an instant message like system like Jabber or something. Either way, I think that's a needed part of the "suite" you're proposing.
__________________
|
|
2008-01-09, 13:54 | Link #18 | |||
Aegisub dev
IT Support
Join Date: Dec 2004
Location: Florianópolis, Brazil, Pale Blue Dot
Age: 38
|
Quote:
And that still doesn't touch the most important point - replicating all of Aegisub functionality in flash would be daunting, not to mention a waste of time... And, I don't know about you, but I'd rather do work on a "real" program. Quote:
Quote:
In the end, "reinventing the wheel" would be the easiest and most effective solution. |
|||
2008-01-09, 14:11 | Link #19 |
My E-Penis > Your E-Penis
Fansubber
Join Date: Feb 2006
Location: Brussels, Belgium
Age: 38
|
i do agree that it sounds interesting. didn't want to diss it so badly. it might work out, but it will be hard to balance out. i'd advise you to gather some experienced subbers and try to brainstorm a bit.
i just meant that personally i wouldn't be able to use it very well in menclave, since we usually do things at different hours, and some people are too reluctant to try out new things. @dj_terk, letting multiple QCs edit the script is a no-no, imho. if it's just 1 excellent QC, that's fine. but good QCs are extremely rare. as i see it, this function would only be useful to get TL/Timing/Editing done quicker. @amz "The system won't allow people to change each other's work. This will be constrained so that you are only capable of editing what you are SUPPOSED to edit. Of course, people being able to work at the same time is the biggest issue." well, tls, editors, and timers all have to have access to the same things really. even if a tl approved a line, he might see the need to change it later, or do a second run. i mean, the main "function" would be the possibility to work simultaneously, which is also its biggest fault. otherwise it's just the same as SVN not to mention that if you use this to get work done faster, you might end up doing it slower. not everybody works at the same rate, so in the end, some people would have to end up wasting more time keeping an eye on the changes than they would originally. oh well, perhaps i just don't understand what you mean
__________________
|
2008-01-09, 14:11 | Link #20 | |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Quote:
__________________
|
|
|
|