Scene timing: Can someone explain it in a nutshell?

[Starks]

Topic.


Additionally, is anyone here exclusively an audio timer?

[Harukalover]

Scene timing: The art of timing a subtitle to end at the exact frame before a scene change occurs. (also start at a change if possible instead of coming in a frame before the scene change).

This tends to have a much more aesthetic look than a subtitle that ends a few frames after a scene change or even a frame before it. The real skill comes in, when trying to determine what to do with a line that has audio that goes a few frames past a scene change.

For these cases you have to decide whether to cut the line off or let it go past the scene change just so it can stay with audio.

Personally, I usually cut it off if the audio for that line doesn't go much further than the scene change. Though I won't cut it off if I believe the line was up for too short of a time to be ended at that exact point.

EDIT: As Quarkboy said, scene timing also affects the starting of a line as well. Starting a line on the same frame as a change or after it usually looks better than starting one a frame before a scene change happens.

[Quarkboy]

Originally Posted by Harukalover

The real skill comes in, when trying to determine what to do with a line that has audio that goes a few frames past a scene change.

For these cases you have to decide whether to cut the line off or let it go past the scene change just so it can stay with audio.

A lot of the art there comes, at least for me, depending upon which syllable is being extended past the scene. If it's a long vowel like "so|u...", then it's less noticeable to cut it off than if there is a consonant after the scene change like "deshi|ta"

Also, sometimes people will purposefully add extra lead-out to a sub that they don't want to cut off, in order to make it look less visually jarring.

Sometimes scene timing also can mean modifying the START of a subtitle too, although generally it's a lot less of a problem.

[endy]

Scene timing is essential to prevent the kind of "flashing" that occurs when the dialogue finishes just before a scene change too. Its all part of fine timing

[Lost]

Originally Posted by Quarkboy

Also, sometimes people will purposefully add extra lead-out to a sub that they don't want to cut off, in order to make it look less visually jarring.

To add to that, a long lead-out (time from when the vocal stops to the time the subtitle disappears/or is replaced) is much preferred to a long lead-in (appearance of the subtitle to vocal start) because while a reasonably long lead-out (I've seen some close to a second) can be tolerated, long lead-ins look very weird.

I was taught to join the subtitle to the scene change (either scene-start or scene-end) if there is less than or about .3 seconds (300ms) between them; and link consecutive subtitles to each other if there is a less than or around a .5 second (500ms) period between them. So far, that has worked well for me [in my practice that is].

At the end of the day, it's up to the timer to QC his timed script to catch bleeds, flashing and/or other awkward visuals.

[False Dawn]

Yeah, the system I work by: if I ever "notice" the timing of my lines, then something's wrong (unless it's alpha timing, which is more flamboyant). After all, people only notice the timing if it's bad, hardly ever if it's good.

Scene timing is very important to stop people noticing the timing. Where it ends near a scene change, it needs to snap out immediately with the scene so that it doesn't create some kind of ghostly subtitle after-image. Same with the beginning of subs - sometimes you have to reduce your usual "lead-in" of lines so that they don't start before a scene change occurs, because again it looks silly.

The visual effect of timing on the subtitling is all important, which is why although most of us audio-time, we usually have the video open alongside to check that the aesthetics work. As Lost says, a timing sweep should be more like a QC of the timing than anything else.

[ffdshow]

Originally Posted by Starks

Topic.

Additionally, is anyone here exclusively an audio timer?

Probably I am. I found that only hearing the audio, click on keyframes and postprocessing would be enough. This is very nice when you are busy.

Unlike translations or edit, not many people actually care unless it's done badly. Do it fast without bleeding and it would be fine.

When the video and audio isn't exactly sync, it's better to do some timing adjustment, although I don't really know how to do it well. Even sync the mouth movement doesn't really work.

[Quarkboy]

Originally Posted by ffdshow

When the video and audio isn't exactly sync, it's better to do some timing adjustment, although I don't really know how to do it well. Even sync the mouth movement doesn't really work.

As far as I'm concerned, if the video and audio are out of sync, that's the ENCODER's problem, not the timers. To fix it, the best way is not to time to mouth flaps, but to foley sounds like doors closing, or even better, to combined video/audio cuts (i.e. say one scene is in the engine room of a submarine with a hum in the background, and the next is in a quiet field. Generally the background sound will cut out EXACTLY at the scene change and you can use this to set the timing offset.)
Of course, if it's more than a linear shift such as a misencoded variable bitrate mp3 or a botched IVTC job, synching the audio is impossible without careful, painstaking manipulation, and it's generally easier to simply find a different raw.

[TheFluff]

If you define "audio timer" as "someone who doesn't view the video while timing" then I guess I could fit that. I use the audio and display keyframes on it, then rely on automatic means to scenetime. Works rather well, actually.

[endy]

I "audio time" as you say except for when i'm just about ready to submit the script, which is when I usually take a look at the episode with the subs on them, keeping an eye out for mistakes.

[Lost]

To be honest, I don't really see how anyone can audio time exclusively. We all audio time, but even if we do not scene time immediately after, I'm sure we all QC it. If a defect is spotted somewhere; to correct it, surely that would involve some sort scene-timing? Please correct me if I'm wrong.

I use post-processing (by about 2-3 frames) to catch frames I might have missed; not usually to replace scene-timing.

[ffdshow]

Originally Posted by Lost

...to correct it, surely that would involve some sort scene-timing? Please correct me if I'm wrong.

It is usually not needed. Those small corrections is not longer timer's work. I only do that when I am also the TSer and encoder.

One question, does anyone time before translation?

[cyth]

Originally Posted by ffdshow

It is usually not needed. Those small corrections is not longer timer's work.

Correcting scene timing shouldn't be the QC' responsibility. Any timer can save them thought on one more aspect of the subtitles with self-checking their work. I really don't see it as a hassle to anyone because it takes just ~30 minutes more, and this way you actually get to see the anime you work on. That's why I love timing; you get to see the anime translated second to the translator, usually with the subs being more literal, and you get to see them before the majority of leechers in a fairly relaxed manner, may I add. Some fansubbing positions usually don't use or even have that luxury (like AFX typesetters).
I don't match lines to keyframes on the audio graph, so doing this pass is obligatory for me, better yet, it's my religion. I'm not wasting my time with it because I rather save on some aspects of the audio timing pass, I get to check and fix my work, enjoy the anime, and save the QC a part of checking (granted they have to trust you).
If your team is not on a strict speedsubbing schedule (one I have yet to witness), then those ~30 minutes of extra work won't be justified as a team loss.

One question, does anyone time before translation?

I'm sure there are people who have done it, but they need a fragging good sense for structures of Japanese sentences, or even better, know Japanese. _yeah did it, but he's a translator too.

[Quarkboy]

Originally Posted by ffdshow

One question, does anyone time before translation?

I time and translate simultaneously for Yes! Pretty Cure 5. I.e. the process is: watch the show once, then make timingraw, load up video and audio into sabbu, and time/translate directly in there. (you could obviously use aegissub or whatever, I just prefer sabbu ).

Timing prior to translation is really only done if, say, the translator is being REALLY slow but the timer really REALLY wants to get things done. I find getting translators to modify .ass files is a chore, though.

[magz]

1) Load the timed script into aegisub.
2) Load up the video/audio/timecodes.
3) Post-process!
4) ???
5) Profit!

At least... that's how I do it...

Oh, and SSA mode FTW.

[False Dawn]

No timing sweep? That puts a lot of pressure on the QC... :s

[TheFluff]

Actually the timing postprocessor with appropriate settings and a properly encoded workraw can be surprisingly accurate, especially if you adjust your audio timing to account for its quirks. I only very rarely need to fix stuff after it.

[Starks]

Originally Posted by magz

1) Load the timed script into aegisub.
2) Load up the video/audio/timecodes.
3) Post-process!
4) ???
5) Profit!

At least... that's how I do it...

Can you be more specific as to how I do this?

[[P]ako]

Tools ----> Timing Post-Processor. Hover around with the mouse pointer for additional help.

[ffdshow]

Originally Posted by False Dawn

No timing sweep? That puts a lot of pressure on the QC... :s

Maybe you are worrying too much. It shouldn't be difficult to time an episode in 2 hours without scene bleed. The problem is still in understanding Japanese language.

As Toua say, there are some advantage as a timer. I get to enjoy the script before others in more original form. For me, it is also a rather funny way to learn Japanese and English.


Another question is... How long does it take to time a script? SSA mode or aegisub mode?

I have switched to SSA mode, It takes 67-90min to time a 340 lines script of simple translations. The funny thing is, an Italian monk seems to be able to time that with half of the time.