I was reading this paper the other day. Which seems to validate what I already know from experience.
We at Viki have a self-imposed limit of two full lines of subs, but other platforms have a very small character limit for each subtitle (resulting in condensed sentences and skipping details). With the excuse that people can’t read fast enough.
But this view has been challenged. Either it wasn’t true to begin with or viewers have become more used to subtitles, and their eyes have been trained to read faster.
I was reading this paper the other day. Which seems to validate what I already know from experience.
definitely depends on amount of time spent reading, fluency with language and general experience reading subs (maybe age?)… because a lot of us can read TCs and subs and still keep up with the feel of the story most of the time
Surely! There are some countries where people are not accustomed to foreign content at all. I think the US is one of them. When there was a film of great success in Europe, rather than bringing it to their viewers with subtitles, they preferred to buy the rights and make a new, American version. One example I can remember is the comedy Three Men and a Cradle (Trois hommes et un couffin), but there are many others. Nikita was remade as Point of No Return, La Totale became True Lies, La Jetée became Twelve Monkeys, La Cage aux Folles became The Birdcage and so on and so forth.
Italian and French people are also not happy with subtitles, because these countries always used dubbing for foreign films and they still do, although in the last decade or so some theatres also offer the option original language option, with subs.
I think it’s like how some people read really quickly and others are just slower because some people read more than others. The more you read, the quicker you’ll read.
As a dyslexic person, and typically a slower reader than a neurotypical person, I can usually keep up with the subs here on Viki, there are a few dramas, usually Chinese, where they speak so fast that I have to pause a lot to keep up. I often rewatch dramas, and it’s then that I catch more of the emotion, instead of just the words.
I think that’s also because we don’t segment the same way and so the subtitle duration will differ from 1 to another platform.
On Netflix, the subtitle duration on screen lasts less than Viki and a single-line sub is preferred over a 2-line sub. From this, they might use less seg combination and have to compress subtitles. The number of characters is limited by the segment duration.
- Include as much of the original content as possible.
- Do not simplify or water down the original dialogue.
- Truncating the original dialogue should be limited to instances where reading speed and synchronicity to the audio are an issue.
- When editing for reading speed, favor text reduction, deletion and condensing but do not paraphrase.
I.2. Character Limitation
- 42 characters per line
4. Line Treatment
2 lines maximum
Text should usually be kept to one line, unless it exceeds the character limitation. Follow these basic principles when the text has to be broken into 2 lines:
– The line should be broken
after punctuation marks
– The line break should not separate
a noun from an article
a noun from an adjective
a first name from a last name
a verb from a subject pronoun
a prepositional verb from its preposition
a verb from an auxiliary, reflexive pronoun or negation
- All subtitles should be center justified and placed at either the top or bottom of the screen"
Last year, I watched more content on Netflix than on Viki (US series, Asian dramas, movies…). We can tell the difference between segging and subbing when switching to one or another platform. Netflix subs have less sub duration (ex with extensions) but shorter subs that are centered.
Their method of calculating the sub duration and whether it needs compressing is different: I think it’s linked to the cps. Why 42 characters max per line?
The number of cps (characters we can read per second spaces included) is usually 21 characters max for the Latin alphabet (it’s about the same cps max in the research paper, 20 cps).
From this general reading speed, I think Netflix and TED created their own set of rules.
Max Netflix: 42 characters / line so 84 characters max for a 2-line subtitle.
1st line = 21 char
2nd line = 21 char
Total = 42 char / 2 sec
Previously on TED subtitling tool:
They have precise counters and if it goes beyond the standard rule, the editor will change the sub.
Segmentation could be modified for 1 language without having an effect on the segmentation of another language because from one language to another, the number of characters is not the same (not the case on Viki, 1 segmentation for all languages). The synchronisation of subs has also different rules.
On Viki, there is no cps or the thing that would relate the most to this cps concept is the length priority rule (which you learn at the correction of your nssa exam and not before it? I don’t remember well for this one).
Nssa guidelines suggest an ideal segment of 3.5 seconds = 7-9 words when possible. But I couldn’t compare “words per second” with “characters per second” (one word could be one or more letters).
On Viki, we don’t share the same combination of rules. We have a 2-line sub more frequently, but play on combining and extending.
A combination of rules helps in itself each rule of this combination. I think that’s the case for Viki and Netflix.
I hadn’t watched for a long time on Viki so when I watched another drama there, it was pretty wordy and sometimes really fast for a 2-line sub compared to what I am used on Netflix (centered especially).
I got to rewind (2 or 3 times) for the 1st ep, but for later ep on, I adopted back the Viki rythm and no need to rewind anymore. We get used to it.
When a 2-line sub appeared on Viki, mentally my mind thought I got to read fast before it disappears so…
I think it’s also a question of a constant rythm of subs appearance on screen. It might be more constant on Netflix because they count the characters per line and centering helps with reading.
Netflix segmentation though is not understandable for me, I can’t find the guide I found the other day.
Here it is:
We can compare with Netflix cps and Viki cps, but we got to download some files and calculate.
Some history on our suggested “two line” limit and the issue of 42 characters maximum on a line.
I have been concerned with the issue of the ideal time and the ideal number of characters for subtitles for a long long time. In the past I looked often at BBC guidelines https://bbc.github.io/subtitle-guidelines/#Timing
The European Association for Screen Translation https://translationjournal.net/journal/04stndrd.htm
as well as the more recent Netflix guidelines. (Netflix guidelines used to say OST MUST be translated but I don’t see that happening regularly. )
The 42 characters standard is not new – it’s been espoused a lot and as I recall there was some talk that it was originally based on the reading speed of a third grader. So there have been some recent calls to increase the number of characters to match the reading speed of the average adult. I think the character limit also has something to do with avoiding having to move the eyes across the movie theater screen as opposed to the much smaller screens we all deal with now. Crunchyroll and dramafever required their subbers to stay within either a 42 character or 45 character limit (I can’t remember precisely).
Two lines. I think all of these guidelines all say a maximum of two lines. If you think about a movie theater screen, the standard was chosen so the audience would’nt have to move their eyes much up and down the screen.
At viki. When viki could first be watched on the 3g iphone (I was the first beta tester) we were concerned that if subbers used breaks and the software also wrapped the text for the small phone screen, there would be five lines of text covering half the screen. To minimize this happening, we eliminated using breaks within the same person’s dialogue in a segment and limited subtitles to two lines. (Today we have larger phone screens but the roku platform is used by a huge percentage of viewers and it doesn’t handle breaks well)
I particularly like the conclusion of this BBC Research & Development Study https://www.bbc.co.uk/rd/blog/2015-09-how-fast-should-subtitles-be “This study has shown that as long as the subtitles are of good quality and match the speech, rate isn’t an issue.”
For both Ninja Academy and Seg 101 we have looked at an ideal segment length of around 4 seconds. Why? Because over time we found that if the person is actually talking for the entire four seconds in Chinese or Korean the subtitle will fit in two lines.
If you read the various organization’s guidelines you will also see that they say to avoid having the subtitle still on the screen when the speech has stopped but we teach the segmenters, if the video allows, to add another second or 1.2 seconds. Why? To give the viewer more time to read viki subs which often are longer than the subtitle for the same dialogue on another streaming site. Are viki subbers verbose? No, they are subbing completely while other sites are trying to stick to the 42 character limit they drop background information, modifying clauses, the full name of an organization or character etc.
This is very interesting information. As a lay person who has read subtitles almost my whole life, I will give you what I think. Reading experience/ability reading is key. I read very fast so English is no problem for me. The only time I stumble on English subtitles is when they are awkwardly constructed.
I prefer to read subtitles over dubbing. (Dubbing is so whack.) But I have always watched international films. Some of those early dubbed voices were terrible!
I think one problem may be subtitles that have a lot of information but the segment is too short in duration. I know the English subber is just listening to the segment and writing the translation but not necessarily paying attention to the segment’s duration. Me? I’d rather have the correct info in the sub and replay if needed.
I like the technical notes that explain background info. I am not afraid of the pause button. So if there is a lot of talk with the action. I will just watch it, then rewind and read.
Reading Hangul - I’ve been learning Korean for awhile now. At first, I had to stop and sound out the Hangul but now I can read it a lot faster. And, now I can read the fancy Hangul script titles too. That fancy script on “Guardian: The Lonely and Great God” that I originally thought was just pretty art says, 도깨비 dokkaebi (goblin) I like the fact that Viki translates signs, etc. I get my Hangul reading practice in that way. But reading Chinese characters are beyond me at this time. I am illiterate - I only know a few characters.
Unfortunately, that is true. There is a lot of ‘English only’ mindset here. But they aren’t watching Viki anyway so we don’t have to cater subtitles to that group. I know some people who hate subtitles because they don’t like taking their eyes off the action to read. And if they are not able to read fast, I can understand their reasoning.
I remember when I graduated, I bothered my panelist (who is super nice btw) about this ideal length, because when hearing and segmenting, I couldn’t always say for sure how many words the subtitle will make, unless I knew a little bit the language.
Then we talked about the length priority order and I always wondered how it was calculated (in general, + 1-3 words = +1 sec to add).
That kind of goes with what the BBC is saying:
“Based on the recommended rate of 160-180 words per minute, you should aim to leave a subtitle on screen for a minimum period of around 0.3 seconds per word (e.g. 1.2 seconds for a 4-word subtitle). However, timings are ultimately an editorial decision that depends on other considerations, such as the speed of speech, text editing and shot synchronisation.”
=> 0.3 x 3 words = +1 sec
Agree with updating the reading speed, I don’t know if there were recent studies to update the reading speed of people on screen (the 42 cps study is from when?). More subtitles are made available and more exposure too.
I know, dubbing always kinda puts me off the show because I keep noticing that their lips aren’t moving in sync. But I do understand those who are just not comfortable reading fast and so like dubs. The only dubbed show I watched willingly was the cartoon Doraemon, and you still cannot convince me that Doraemon speaks anything other than Hindi
basically I think dubbing removes a part of the taste of culture you get from foreign shows
I would like to contribute a more recent (2021) study on the speed of reading subtitles which I think is relevant to our discussion.
And here, for whoever is interested, are the N F captioning and subtitle requirements:
Thanks for the article. I see that it’s from a Sydney University Nice! I like the insight that this paper offers.
As you know, I’ve done a bit of study into what various platforms define as ‘good’ subtitles from a character-per-second perspective. What some studies don’t take into account is that the attention of viewers is pulled in three directions. These are: (a) time to read the subtitle; (b) time to keep abreast of the on-screen content and action; (c) time to hear and interpret vocal tones such as anger, fear, happiness, etc. I haven’t yet found a study that considers all these aspects. Most studies seem to concentrate on reading the words. I have a feeling I read something once about how to share words with on-screen content in a balanced way. I’ll have to see if I can find that paper. It might be and interesting comparison to this one that you’ve posted.
Anyway, I digress. Thank you for this insight. I appreciate the psychologists’ analyses and their scientific method. I need to work through that in a little more detail, now.
Oh yes, do share if you find it, I’m sure it will be very interesting.
Even thought I’m fluent in english, I read noticably faster in swedish, which is my native tongue. I watch mostly historical c-dramas. Two things I’ve noticed here on viki are: The chinese speak very fast, faster than I can read english subs That is because the subs disappear when they stop talking, even if there is noone answering and the subs could have stayed for a second or two longer. Some subbers are kind and explain things that otherwise would have been lost in translation and then you really need to hurry to keep up.
I usually don’t look for swedish subs because there usually insn’t any and even if there were, translating yet again and there would be even more that’s lost in translation.
I think Chinese languages are very succinct, dense. They say a lot with very few “syllables”. Some period dramas also use proverbs instead of normal phrases and that makes it even harder. A 3-4 syllable proverb/expression needs a full if not even long and complicated sentence in English and other Western languages.
This is just bad segmenting. Viki paid segmenters tend to do that. They routinely cut just after the actor has finished talking. The NSSA segmenting school always tells us to leave enough time whenever possible, the general rule being 1 second extension, but in dense subs with people talking quickly, more than that.
Maybe in that show you watched Quality Controls by volunteers hadn’t been done yet or at all (or the segmenters happened to not remember what they were taught in NSSA).
I agree with this. I usually read slowly because I have a problem with my eyesight, but when I watch Kdrama I can read the subtitle fast. Maybe because my eyes are trained to read faster.
Hi again, Irmar.
Here is one of the articles about which I was thinking. The researchers have studied three speeds (12, 16, 20 characters per second).
From the abstract…
We measured viewers’ comprehension, self-reported cognitive load, scene and subtitle recognition, preferences and enjoyment. By analyzing people’s eye gaze, we were able to discover that most viewers could read the subtitles as well as follow the images, coping well even with fast subtitle speeds.
UPDATE: I’ve just discovered that this article is the one that Irmar posted at the start of this thread. I’m leaving the link here so that it makes the article easier to find now that it is two years later.
Isn’t it the one that I linked on the very top of this thread?
I thought this was the general rule; to stop the sub when the person stopped talking. But then I hope more will use the quitetime to extend the subs when needed.