Timing Reviews – A Brief Into Timing

Dark_Sage edit: I won’t be doing the timing reviews. Revy-chan will. Note the author of this post. Okay, onto the content.


Timing Reviews

Timing is a simple job. It’s taking the text given to you by the translator (usually either in .txt or .srt form) and then attaching them at appropriate times with dialogue. All though it is essentially a simple task, there are still people who manage to fuck it up beyond all understanding. I’m not the best timer in the world, but I do carry a fairly high standard when I do my timing. There is something that needs to be said beforehand however, and that is that a lot of timing comes down to experience and personal feeling. What I mean by this is whether you should snap text to key-frames, how long your lead-ins and lead-outs should be, differ on a personal basis. I’ll explain what I feel is best as well as why; and that will be the how I review groups on their timing. Aside from this however, scene bleeds are just a big EPIC FAIL and any groups that leave scene bleeds in are idiots. They’re not hard to spot, and it if passes the timer (it happens, even with me) and however many rounds of QC you do (should be more than 1) I will judge you even more harshly than adding a 500ms lead-in/lead-out for key-frame snapping. Especially if your scene bleed is more than 1 frame early or extended.

Timing Terminology

Key-frame: A key-frame is the start and end of a smooth animation. That is, points after one block of animation surrounded by key-frames are generally a transition in scenes. Some tend to be less subtle (ie. screen pans upwards) while some tend to be entire changes.

Lead-in: The amount of time added to show the subtitles before audio is actually heard. Most people will add a lead-in of anywhere between 100-200ms. Adding no lead-in is just herp derp stupid.

Lead-out: Same as a lead-in except it’s on the opposite end. Time added onto the end of subtitles so that they linger on. Lead-outs are generally much longer than lead-ins and are usually anywhere between 200-400ms. As with lead-ins, having no lead-outs is just herp derp.

Scene-bleeds: Ah, scene-bleeds. These are the worst things that you can leave in as a timer. Essentially, scene-bleeds are where timers fail and make subtitles appear too early, not early enough, end too early or not end early enough when near key-frames. I think it’s best to use some visual examples to explain this.


In the above example, the line ended too late and bled onto the next scene.


In this example, the line came on too early and bled in from the previous scene. This one actually would’ve been fine if a fade-in was added to fit in with the effect on the television.

Line-linking: This is essentially linking lines that are within close proximity to each other to eliminate the “blinking” of subs going on and off the screen. This can link into scene-bleeds in the sense that a line is close enough to the start of a new key-frame but the timer doesn’t snap to the key-frame and the sub “flashes” onto the screen after a scene change.

My Standards

Aside from scene-bleeds being a big NO, my usual lead-ins are 140ms, while lead-outs are 300ms. I will link to key-frames with lead-in of up to 250ms and lead-outs of up to 400ms. I personally feel anything longer than that just feels really awkward, especially excessively long lead-ins. Sometimes I might end lines earlier if they’re too far from the ending key-frame to make it not blink off the screen, then have a scene change. These numbers are just for reference though. I do sometimes add longer lead-ins and lead-outs if I feel it works with the scene.

How The Reviews Will Work

There are pretty much only 2 sections when it comes to reviewing the timing of a show. They are:

  1. Scene bleeds
  2. Flow

Under scene bleeds, I will note down and snapshot each scene bleed and how many frames behind/extended they are. As for flow, that will cover how well lines link together (no blinking subs), lead-ins and lead-outs near key-frames and whether they were excessively long or too short. Please note, I will judge MUCH MUCH harsher on scene bleeds as I stated above. Having more than 2 scene bleeds, especially if they are over 2 frames in length will pretty much take you down into a B grade zone. This is a new feature so there may very well be some hiccups along the way, but hopefully I will reach Sage-kun’s level of criticism quickly.


Forgot to mention that the timing reviews will be added as an ADDENDUM to D_S’s reviews. I’ll be starting with Ozma and adding them to the bottom of D_S reviews when done. So do check back for timing reviews. THEY WON’T BE A SEPARATE POST. I will update the categories as needed once the review has been added.

0 thoughts on “Timing Reviews – A Brief Into Timing”

  1. How much of your flow score will be subjective? For example, I prefer leadout to be more like 400-500ms and either linked to following lines or snapped to the next keyframe whenever possible as I don’t really like the look of subtitles blinking on and off all the time during dialogue. I think things like that are pretty much up to the timer’s preference/style.

    There’s not much excuse for scene bleeds or more than about 300ms of lead-in though, that’s quite true.

    • As subjective as needed. I mentioned that a lot of it is personal preference/style, so as long as it’s not overly retarded, I won’t mention it. I will mention the really bad things though. I’ll have a better idea once I actually do a few reviews and get a feel for how I should grade everything so the first few reviews might not come for a bit.

    • It’ll be hard, which is why I’ll only be pointing out things that really stand out to me as I noted above. I’m not gonna fuss about the minor details. I’ll likely change the grading system once I actually do a few reviews and find a way to grade them in a way that makes sense.

    • If I reviewed timing I could probably get it down to five grades or so:

      One star – Timer clearly has no idea what they are doing and has ignored scene changes and the general principles of making the subs actually readable. Almost unwatchably bad, even to non-timers.

      Two stars – Timer has a grasp of the principles of timing but due to carelessness or other technical issues there are many cases of scene bleeds and mistimings that would drive anyone insane if they know what to look for. Watchable, but bad. Most CR stuff I’ve seen falls into this category.

      Three stars – Timed competently, though without a great deal of care. Most lines are fine, but there may be a few scene bleeds or other errors here and there throughout the script.

      Four stars – Timed well, with at most one or two very minor bleeds (no more than 1-2 frames) and nothing particularly obvious. Some attention may have been paid to line splitting, fading, etc but less than the next class.

      Five stars – Timed almost perfectly, with at most one bleed and particular attention paid to correctly splitting lines for proper flow and dramatic effect. The length of the onscreen text matches the length of the spoken lines as closely as can be managed. Fades used appropriately on subtitles that end during crossfade scene transitions.

      Or something like that, anyway. I guess my main point is that there’s a lot more to consider between ‘acceptable timing’ and ‘really good timing’.

  2. Am I the only one who doesn’t get the point of screenshoting scenebleeds? Those two screenshots don’t tell me about anything but the styling being quite bad and I don’t think it’s actually possible to “show” scenebleeds like this.
    If you really want to include some screenshots – IMHO aegisub audio view with keyframes loaded would be better. Something like this: http://arinashi.com/img/gslsdiwa.png
    Obviously it would require more effort, but any timer can say that there’s a scenebleed just buy looking at the screenshot now.

    • Key-frames aren’t always correct. That’s probably one of the misconceptions that lead to scene bleeds in the first place.

    • No I agree, there’s no real point in taking screenshots of timing issues because the whole point of timing is that you can only spot problems with it by watching the video in motion. Usually if I’m checking timing for someone else and I want to report errors I just paste the offending lines from the .ass file, so maybe the same could be done here.

      Those screenshots just look to me like the default Aegisub style, so I assume the scripts were timed without styles or something (not really recommended since you can’t spot 3 liners)

      • Well usually I’d do the same. I’d note down the time where the scene bleed occurred and how many frames. People like pictures though and if I just wrote 2-3 paragraphs on errors it might not be as enjoyable to some people. But yeah I do understand that those pictures above don’t really show anything aside from the default stylings. I’ll have to think about how to present the errors I find still.

  3. Isn’t there some sort of WP plugin or something that would only animate the gifs when I click on them? You know, sort of like on the ‘chans? I think the gifs are a good idea, just would be nice if I could activate them one at a time.

  4. This should be interesting. I don’t think anybody has really checked my timing since a month or so after I joined the group. Haven’t had any complaints though. It’ll be interesting to see what you have to say.

  5. Timing is…
    not important.

    Timer is…
    no body.

    A timer messes up when he…
    fapped in the middle of timing (or zzz)

    The value of a timer is on…
    whether he can tl/encode/tlc/edit/QC, since no one cares about timing.

    /me runs.

  6. I can look past many things in a fansub, but subpar timing is one thing that’s terribly distracting to me (probably because it’s the only part of fansubbing I can do decently, ha ha). So yeah, relevant to my interests etc.

  7. I don’t see much point in reviewing timing, something that takes zero skill. The quality of timing just depends on the amount of time you take, assuming you know what you’re doing. Actually making GIFs of scene bleeds seems like a god honest waste of time.

    • If timing took no skill then there wouldn’t be some timers who can do a better job in an hour that other timers can do in 3-4.

      There also wouldn’t be so many groups who just have consistently awful timing in every release.

      • No, timing doesn’t take skill. None at all. The reason some timers can do a better job more quickly than others is because they’ve probably done it longer, learned all the hotkeys and shortcuts in aegisub, and care about quality. In essence, they know what they’re doing while others don’t. Anyone can learn how to time properly and quickly in a matter of hours. Something that can be mastered so easily doesn’t take any skill.

        • That’s like saying playing [insert game here] takes no skill because I’ve played it longer than others and know all the hotkeys and shortcuts in [insert game here] and care about winning.

        • >The reason some timers can do a better job more quickly than others is because they’ve probably done it longer

          I’m not sure what your definition of the word is, but I’m pretty sure in English this is what we’d call “skill”.

          “the ability, coming from one’s knowledge, practice, aptitude, etc., to do something well”

        • In fact, let’s just flip that around to the other roles:

          Translating takes no skill. The reason some translators can do a better job more quickly than others is because they’ve probably done it longer, learned all the words and phrases in Japanese, and care about quality.

          Editing takes no skill. The reason some editors can do a better job more quickly is because they’ve probably done it longer, learned about the correct rules of English grammar, and care about quality.

          Typesetting takes no skill. The reason some typesetters can do a better job more quickly is because they’ve probably done it longer, learned about all the useful little tricks in the .ass subtitle format, and care about quality.

          Encoding takes no skill. The reason some encoders can do a better job more quickly is because they’ve probably done it longer, learned about the correct filters and settings to use in certain situations, and care about quality.

          In short, fansubbing takes no skill and you might as well throw together a group out of random people from IRC with no experience and have them sub everything. And why bother reviewing fansubs, because it takes no skill and everyone’s subs are the same. You should probably just join Hadena since I think you guys are on the same wavelength here.

          Also, typing giant walls of text takes no skill, it just requires a blog that entirely ignores paragraph breaks :(

          • Okay, right. You need a line break, then ampersand nbsp semicolon, then another line break. I dunno what the deal is with that, but it works.

            • Maybe it’s just different for me because I’m always logged in as admin, but I just press the enter/return button twice and it creates a space between the lines.

          • Yeah, that hasn’t worked for me since you switched to the new version of WordPress. Even if I add paragraph breaks, they all get taken out as above.

          • You both conveniently didn’t respond to the part of my post that matters, so I’ll paste it here again for you: “Anyone can learn how to time properly and quickly in a matter of hours. Something that can be mastered so easily doesn’t take any skill.” Since you want argue using analogies, let me use one that makes sense: Some people can dig deep holes faster than others. Digging holes requires nothing but a shovel and a pair of arms. Oh, but wait, ALL activities take skill according to you, so let’s not forget the SKILL it requires to dig that hole! Right? No. There are reasons things like “ditch digging” and “burger flipping” are called UNSKILLED JOBS. However, based on what you’ve said so far, you would disagree because you think it takes skill to push a shovel into the ground or a spatula underneath a burger simply because some people can dig holes and flip burgers faster than others. On the other hand, editing, encoding, typesetting, and translating all require advanced knowledge, and depending on how smart you, are a few weeks to several years of experience to do them well. Therefore, they are SKILLED JOBS. Timing takes at most a day or two to master. Something that takes such a short amount of time to MASTER is not an activity that requires SKILL. Get it? Also, you seem pretty upset. I guess you must be a timer? Keep in mind that I’m not saying I don’t appreciate the work of timers. I really do. I also appreciate the guy who made this Subway sandwich I’m eating. However, just like timing, making a sandwich is not a skilled job.

          • I’m an editor who also does timing, typesetting and QC on the side.

            And no, not really. If timing was so easy to master we wouldn’t have so many groups that are so terrible at it.

            A fast, accurate timer is a valuable asset to any fansub group, and you can’t just train those overnight. If you disagree, try timing a full episode sometime and see how you do in review. Hit me up on IRC and I’ll check it for you, if you like.

          • I never said timers aren’t valuable. In fact, if you read my post, which I’m almost certain you didn’t, you would see that I said I really appreciate timers. Not sure why you asked me to time an episode for you to review. That would be a grand waste of time. Sorry, but nothing you’ve said so far has made any sense. So many groups are terrible at timing because they don’t care about quality. Anyone can master timing easily, but if they don’t give a shit about quality, then the outcome is obvious. If you still think timing takes skill, you’re simply in denial.

          • Timing is a skill. Thus, it takes skill. Editing, translating, and encoding are the only jobs in fansubbing that require zero skill; they require knowledge. I can’t believe this is so hard for people (namely you, Janice) to understand.

          • >Not sure why you asked me to time an episode for you to review. That would be a grand waste of time.


            I was hoping a little bit of first-hand experience might do something to cure the rather dire case of ignorance you have there, but at this point I’m thinking you may be beyond help.

          • Oh, my God. Fine, then. Timing takes skill. VERY LITTLE skill. Just like the other unskilled jobs I referred to in my previous post. Pressing buttons on a cash register takes skill – VERY LITTLE SKILL. I can’t believe that’s so hard for YOU to understand.

          • Wow then I guess you must have mastered it already! I can’t believe you’re not timing everything for WhyNot considering that after mastering timing you’re presumably able to turn out error-free scripts faster than Sindalf.

          • Given that you’ve mastered timing you should be able to do a whole episode in 15 minutes, so I can’t really imagine you’d have very long to get bored.

          • >I also appreciate the guy who made this Subway sandwich I’m eating. However, just like timing, making a sandwich is not a skilled job.

            Woman, it is clear to me too long has past since you’ve last made a sandwich. Since we’re in a rush, I’m sure Subway will let you borrow their facilities. A beer will do nicely with that. Thank you, m’dear.

          • im just saying it’s worth reviewing

            since, you know, there’s a lot of difference in how well groups do it

            just like editing

            holy shit man

    • I agree about the GIFs though, that kind of thing’s overkill. A brief description of the timing quality in the episode would be fine by me.

    • If timing was so easy, no-one would spend more than 20 mins on an episode and people would be content with a dialogue timed episode with TPP added.

    • I’d say the problem is exactly that many “timers” don’t know what they’re doing. Once you get past that, that’s when it comes down to time spent (only not really, since there are clear differences in the speed/quality factor which separates good timers from decent ones, but yeah).

      I’m not sure I’d call it a matter of “skill”, but the amount of releases with subpar, if not plain shitty timing is enough reason to justify doing this, just no need to go overboard.

      • This is also quite true in the sense that I feel most timers just dialogue time, then add TPP to everything. It’s obvious to see when you get overly long scene extends (like the shit you see in drama timing).

    • For me, timing is only a audio-text matching game. I don’t think this is reviewable. It’s a surprise that someone would take time to do timing review…

        • I think professional subtitles do that? :D

          It would be nice to have subtitles with accurate translation and good English. But I can’t brag to friends, “hey, my subtitle is awesome because the lines start and stop in the scene break.”

  8. I’ll still do the GIFs but instead of placing them in the article, I’ll leave a link so those that care about stuff like that can see. I will still write a changelist of errors.

  9. Having timed around 300 episodes of anime, this guide is needless and unnecessary.

    Timing should be ranked on a 0 or 1 scale.

    0 = lines don’t appear when lines are said.
    1 = lines do appear.

    Seriously though, all this lead in and lead out rules are bullshit. Just make the line appear when the line is said. And who gives a shit about scene bleeding? Most professional subs have them.

    Timing is the 2nd most important task in fansubbing. It should take around 10 minutes for 100 lines. Any more and you’re doing it wrong. Yes it’s monotonous and boring as hell, but every fansub needs it.

    The only time I see bleeds being a problem is if they go into a title scene card.

    Have a nice day from a timer who probably has timed more episodes than most timers here combined.

    • Your point being? You can’t appeal to professional subs as the heralds of timing rules. Professional sub tracks suck because of limitations in the hardware and subtitle format. Advanced Substation Alpha was made to address those shortcomings, so if you choose to be lazy and not use Aegisub well, you’re the one at fault.

  10. What metric are you measuring “suck” of timing? The lines appear when they need to, there’s no holes in the script. What more do you expect?

    Professional subs are for readability, which they accomplish very well.

    Using Aegisub is an indicator of being a good timer? My god what are you people smoking.

  11. I still don’t get why this is turned to a drama, Revy-chan doing for people that care about this, there is no reason for anyone to complain, if you don’t like it then don’t bother looking. Over.
    /me hide behind Dark_Sage.

  12. Worse than any of the things mentioned here is when typesets for on-screen stuff aren’t accurately frame-timed, especially when typesets bleed into other scenes. THAT more than anything drives me up a wall.

  13. @Xythar
    Your loss.

    It’s not exactly drama, more like this topic in itself is quite unnecessary; timing is either right or wrong, there’s not exactly a “standard” that it can be reviewed by.

    • So when your group gets a timing review and gets a big fat F you won’t care? Not likely.

      Also, you’re right, timing isn’t what makes Doremi shit.

    • Your group is widely renowned for its incompetent timing. Perhaps if you were less of a pariah you’d know.

      Then again, people tend to ignore the little issues when there are much bigger ones.

    • If you refer to the Ozma Timing Reviews, I put them under linking errors as to how I’d deal with scene-bleeds with dialogue. Note that this doesn’t always work and it depends on the scene, how long of a bleed, etc.

  14. A serious question about scene bleeds, where the line extends beyond the scene change: when should you cut off the line, and when extend into the next scene? If the line continues on for seconds, it’s obvious, but what if only one word extends beyond the scene change? Or one full syllable? Just looking for an opinion here; no “right” answer.

    • Really depends on the situation. If it’s one syllable, or an extended syllable (like a scream) I’d pretty much always cut it off. If it’s more than that it depends on whether it looks worse to continue the subs or cut them off early.

      I’d say that generally if the scene change cuts to showing something completely different, particularly an action scene or otherwise something where the viewer’s attention will be focused on the onscreen action and not reading the subtitles, it generally looks better to cut it off even if it’s a word or two early.

      If it cuts to another shot of the character talking, or another character reacting to the first character’s line, it’s usually better to let the subs run past the scene transition. If you do this, make sure they extend a good 500ms or more past the scene change regardless of whether the audio finishes earlier because having subs end <500ms after a scene change looks crappy regardless of the circumstances (unless of course they get cut off by another scene change)

      Really though, there's no hard and fast rule. You just have to watch the scene afterwards and go with whatever looks best.

    • Some people use rules like “cut no more than one syllable” or “cut no more than half a syllable.” When I’m timing (qcing to be more accurate since I rarely time from scratch) I use my judgement on any individual line something like that occurs on.


      If I cut the line short and it doesn’t seem like it is cut short than I’ll go with that. Conversely, if I extend the line and it doesn’t seem obtrusive because it crossed a scene change than I’ll go with that. Both might work sometimes and I make a judgement call.


      There are factors that play into what seems right. Softer sounds like “sh” are easier to cut off than harder sounds like “k” since soft sounds sort of “fade out” at the end. Also if there’s a lot of background noise than it’s easy to cut a line short since the dialogue gets lost some in the noise anyway. Same goes for action scenes that distract the viewer. It’s easier to extend if there is dialogue in the next scene to chain to. If you do extend and there is no chaining involved you need a healthy lead out.


      tl;dr: Whatever makes it seem natural during playback.

      • >It’s easier to extend if there is dialogue in the next scene to chain to. If you do extend and there is no chaining involved you need a healthy lead out.

        This is also solid advice.

    • Pretty much what Xythar said. Preferably, anything longer than 200ms though, I won’t extend cause it looks like shit.

  15. My my what a ruckus over such a task.
    Listen to audio (pretime with small lean-in [because it’s confusing as fuck to have a line displayed 400ms before the audio starts and ends too early or 2 seconds too late]), watch video and clear all scenebleeds caused by the encode or yourself (no, not all keyframes are a 100% indication that you have to stop displaying a line – try timing to a BD raw for once), extend/shorten where necessary and avoid using tpp.
    There, you have solid timing in less than one hour.
    Easy as that ヘ(゚∀゚*)ノ


Leave a Comment