You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This issue is a follow-up to the other subtitles-related issue: #22.
I have found that with both auto- and non-auto subtitles it is very often the case that the subtitles are cut in the middle of a sentence and the resulting mp3 sounds very incomplete. It looks like two possible optimizations can be applied in order to cut the cards sentence-based. This can be ensured by merging the subtitles that follow each other closely.
When the subtitles are manually created and well-written, they can be merged together to form full sentences so that the audio files are cut sentence-by-sentence not part-by-part. This can result in a very precise cutting and therefore in a much better user experience out of the box. For automatically generated subtitles, at least a distance check can be made: If the subtitles are very close in time to each other, they can be also merged because it's likely that they are part of the same sentence spoken in the video. Both algorithms could be enabled with dedicated option checkboxes in the UI.
Not automatically generated subtitles
This is a very good example of very well crafted subtitles: https://www.youtube.com/watch?v=GfF2e0vyGM4. Each of the subtitle forms a part of a sentence and at some point the sentence finishes with a good punctuation sign like . or ?. In this case, it can be optimal to concatenate the subtitles until one sentence is full finished. This should make perfect Anki cards, one card per sentence.
Automatically generated subtitles
In this case, it is not possible to rely on punctuation. Instead, one subtitle range can be checked with a previous one:
If the distance between subtitle ranges is less than 0.1s, they can be concatenated as forming the same sentence. Otherwise, they are two different cards and should be kept separate.
This issue is a follow-up to the other subtitles-related issue: #22.
I have found that with both auto- and non-auto subtitles it is very often the case that the subtitles are cut in the middle of a sentence and the resulting mp3 sounds very incomplete. It looks like two possible optimizations can be applied in order to cut the cards sentence-based. This can be ensured by merging the subtitles that follow each other closely.
The original description can be found here: #7.
Not automatically generated subtitles
This is a very good example of very well crafted subtitles: https://www.youtube.com/watch?v=GfF2e0vyGM4. Each of the subtitle forms a part of a sentence and at some point the sentence finishes with a good punctuation sign like
.
or?
. In this case, it can be optimal to concatenate the subtitles until one sentence is full finished. This should make perfect Anki cards, one card per sentence.Automatically generated subtitles
In this case, it is not possible to rely on punctuation. Instead, one subtitle range can be checked with a previous one:
If the distance between subtitle ranges is less than 0.1s, they can be concatenated as forming the same sentence. Otherwise, they are two different cards and should be kept separate.
Using this technique, I could achieve a decent cutting of the sentences as spoken in this video: https://www.youtube.com/watch?v=7I4J4vy2Deo&t=12s.
Both techniques could be activated with a corresponding option in the UI.
The text was updated successfully, but these errors were encountered: