wotreplays.org: Replays to Youtube video conversion: the big picture
This feature-in-development really deserved it’s own post due to the rather large wall of text you’re about to consume…
It’s been often talked about, and has even been pulled off by a Russian site, but the lead time on getting a replay converted to a Youtube video is rather large. I’ve considered running it on my own workstation, but living where I live, a 512kbps upload speed is the best I get. That means that uploading a 1080p video can take up to 24 hours, if not more. And that’s on a good day.
That’s why the initial prototype was set up to run on a server or high-end workstation that had the graphics card horsepower to actually run WoT at a decent detail level, at a resolution big enough to record 1080p video. That prototype didn’t pan out very well due to the limitations of actually recording WoT running in WINE under Linux; the vagaries of OpenGL and the various GPU architectures make it a pain in the ass. Granted, it’s doable, but not that easy.
The second prototype (the one that’ll eventually eat your WoT install) runs on Windows, and uses Fraps or Bandicam to record video. This is the prototype that will get fixed (so it stops eating installs) and will probably be the way things will be done. But how exactly does it work?
Quite simple, actually. All it does is occasionally ask the site’s backend if a replay needs conversion. It then downloads the replay, and uses it’s own private copy of your WoT install to play said replay. A few mods that are added to said private copy will trigger events that can be picked up by the application, which will then do the appropriate voodoo magic to get Fraps to start recording. After the replay finishes, the completed recording is then uploaded to the site’s backend, where it will be post-processed and sent off to Youtube.
Of course, it’s not a very light-weight process. That’s why initially (if it gets off the ground, that is) only one replay will be converted per day; that replay being “the best” of that particular day. How “the best” is figured out is still up in the air, either most likes, most downloads, or just a case of “uncle Scrambled picks it”. Beyond that the big picture idea is that converting a replay is going to cost you money. Not that much, but it will cost. Why? Because that money will be split between the site and the people running the conversion application – and the split will be in favor of the people, just so we have that clear (think 30/70). This does mean that the whole process needs to be made bullet proof in order to prevent people abusing the system, and since money is involved, it can’t “lose” work done, or otherwise make “mistakes” in numbers.
Long story short: it’s in the pipeline, but it will be a while.