chirag127
commented
3 years ago phonetics (& necessary physiology). ~<0.7 seconds each. Some of these pauses make words(/phrases) more recognizable or just familiar (even if their origin would be physiology/hard to avoid physically).
We should skip it .
2. category dependent error. I.e. in Live stream & recordings of lectures.
related to
5. Exclusion/s: Music, .., ... ?
- 0.5s-10s rethorics (or minutes of silence, to express relevance solemnly/diginity/equanimity)
We should skip it .
- All silence-skipping could analyze video too
I don't think it is needed. Frame may change a lot during a segment which is intended to be skipped and hence create too many false positive.
- these can be skipped in inactive tabs more likely meaningfully than in active ones.
we should skip it in both active and inactive ones .
5. Exclusion/s: Music, .., ... ?
Of course, if easily possible. May provide opt-out option for disabling skip silence in music .
Micro silences" appearing everywhere,
May provide opt-in option for enabling micro silence skipping or just do it by default and not provide option to opt-out of it . as I think skip silence is absolute skip silence to save time .
- less complicated feature idea: dynamic speed based on speed of speech (syllables per minute or wpm)
Yeah, it is very easy if there are captions in the video .
- dynamic speed based on speed of speech (syllables per minute or wpm
How will we calculate the WPM in the video that don't have captions ?
ImprovedTube
allanlaal
yay! + another conditional speed/skip based on content? (https://github.com/code4charity/YouTube-Extension/issues/899#issuecomment-846610717 )
[ ] technically simple implemenation: Dynamic speed based on speed of speech (syllables per minute or wpm) & difficulty (average letters per word) (based on transcript & category - not audio)
hi @vantezzen https://github.com/vantezzen/skip-silence (MIT license!)
What kinds of silences are typical?
phonetics ~<0.7 seconds each. Some of these pauses make words(/phrases) more recognizable or just familiar (even if their origin would be physiology/hard to avoid physically. Some phonetics might be side effect of vocal physiology, rather than useful?).
category dependend error. I.e. in Live stream & recordings of lectures. The less scripted the sound / talk was, the more likely there are pauses due to human error, lack of performance/exhaustion of the speaker, combined with 'um'- & 'hm'-sounds (or other errors, like mono #552).
0.5s-10s rethorics (like minute/s of silence, to express relevance solemnly/diginity/equanimity)
[ ] these can be skipped in inactive tabs more likely meaningfully than in active ones.
[ ] in active, visible tabs, all silence-skipping could(?) analyze video too. ( - Video frames should not change much noticably during the whole skipped period)
On 1. : Is all existing (proportional) digital speed-up only skipping random parts? "Micro silences" appearing everywhere, could be removed from every syllable or word in a speech, without changing their proportional tempo. Thus changing how the speaker's energy is perceived, but no changing the clarity of words as much - nor the relational speed (pauses might be semantically relevant. )