Speedreading video
21 October 2025 09:23 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
igenlodeAfter spending so much time referring to YouTube's auto-generated transcripts to try to decipher Russian (they are basically a timestamped listing of all the auto-generated subtitle text, I think, so they enable you to look back at the context of an entire phrase as opposed to just the handful of words visible on the screen at any one time), it dawned on me a couple of days ago that it is of course possible to read the transcripts of *English* text as well. Which is particularly useful in the case of videos which are basically a voiceover with marginally-related images screened for visual appeal, because it is possible simply to read the transcript and not bother to actually listen to the speaker at all, especially if the title appears to be clickbait :-p
It's the equivalent of playing the sound back at 6x speed, without the high-pitched gabbling effect :-) Of course the text isn't formatted into neat sentences and proper names are often weirdly mangled, but it makes it possible to skim through to the actual relevant content, if any, and/or go even faster in order to ascertain that there isn't any...
In Russian, ironically, reading the transcript is actually considerably *slower* than simply listening to the entire video. Partly because I'm generally doing so in order to be able to stop and look up vocabulary, whereas when I'm listening to the audio I'm just taking it for granted that I can't understand a lot of what is flying past, and partly because I can't speed-read Cyrillic (in fact I often have to read it out loud, or at least mouth the words!) I can skim Russian text for names of interest by jumping ahead to the capitalised words, but that's about it.
But also, I rely quite heavily on body language and lip-reading the speakers where possible (voiceover text is harder to follow in that respect, but conversely often better enunciated and, having been pre-prepared, more grammatical!) So understanding an unformatted transcript without the accompanying visual/audio clues as to where sentences/phrases start and finish tends to be very tricky; they are useful for transcribing unknown multi-syllable words and, as mentioned, for providing a wider 'window' into what has just been said, but not really accessible as standalone documents.
And there are often individual words or sections of a sentence that aren't recognised and don't get subtitled, and these don't show up in the overall transcript. So when looking at that in isolation you get a gap in the meaning which isn't highlighted in any way, and makes nonsense of it unless you know that you are looking at two unrelated half-sentences with a missing section in between... It's the universal problem with all foreign content; I may be able to follow it when it is textbook-correct, but the instant somebody misspells a word or uses bad grammar, or a few words of audio drop out, my level of comprehension instantly starts to fall dramatically :-p I just don't have the level of familiarity with what *ought* to be there that would enable me to fill in the missing pieces in the way that I can almost unconsciously do in English.
It's the equivalent of playing the sound back at 6x speed, without the high-pitched gabbling effect :-) Of course the text isn't formatted into neat sentences and proper names are often weirdly mangled, but it makes it possible to skim through to the actual relevant content, if any, and/or go even faster in order to ascertain that there isn't any...
In Russian, ironically, reading the transcript is actually considerably *slower* than simply listening to the entire video. Partly because I'm generally doing so in order to be able to stop and look up vocabulary, whereas when I'm listening to the audio I'm just taking it for granted that I can't understand a lot of what is flying past, and partly because I can't speed-read Cyrillic (in fact I often have to read it out loud, or at least mouth the words!) I can skim Russian text for names of interest by jumping ahead to the capitalised words, but that's about it.
But also, I rely quite heavily on body language and lip-reading the speakers where possible (voiceover text is harder to follow in that respect, but conversely often better enunciated and, having been pre-prepared, more grammatical!) So understanding an unformatted transcript without the accompanying visual/audio clues as to where sentences/phrases start and finish tends to be very tricky; they are useful for transcribing unknown multi-syllable words and, as mentioned, for providing a wider 'window' into what has just been said, but not really accessible as standalone documents.
And there are often individual words or sections of a sentence that aren't recognised and don't get subtitled, and these don't show up in the overall transcript. So when looking at that in isolation you get a gap in the meaning which isn't highlighted in any way, and makes nonsense of it unless you know that you are looking at two unrelated half-sentences with a missing section in between... It's the universal problem with all foreign content; I may be able to follow it when it is textbook-correct, but the instant somebody misspells a word or uses bad grammar, or a few words of audio drop out, my level of comprehension instantly starts to fall dramatically :-p I just don't have the level of familiarity with what *ought* to be there that would enable me to fill in the missing pieces in the way that I can almost unconsciously do in English.
no subject
Date: 2025-10-25 03:06 am (UTC)I use YT's auto-transcripts all the time to text-search for keywords to find spots in long videos that I might actually want to listen to. There are some people in my circles of interest who do these looonnnggg interview videos with various people who are/were in the games industry, and that's neat and all, but if someone worked on 5 different franchises and I'm really only interested in one of them, I do not want to have to be actually listening through 4 hours (even on 1.1x or 1.25x or something) to learn if any of it is of interest to me.
no subject
Date: 2025-10-25 05:30 pm (UTC)