Giters

pietrop / subtitlesComposer-app

A simple os x desktop app built in electron, using Aeneas under the hood to generate captions files from media(audio or video) and plain text transcriptions without time-codes.

Home Page:https://autoedit.gitbook.io/subtitles-composer-app/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Adding input format of type "subtitle"

braincoded opened this issue · comments

commented

Thanks for your question @braincoded

Could you tell me a bit more about the thinking behind this request?

The app uses this module to handle the alignment subtitlesComposer.

Grand part of the modules does a pre-segmentation prep to create a file in that Aeneas subtitles format before calling aeneas.

See here for more details on the pre-segmentation steps involved in creating that Aeneas subtitle file programmatically before calling aeneas

The thinking is that if we can get the presegmentaion as good as possible, programmatically, there might be close to no need to manually edit the captions once Aeneas has done with the alignment (?).

In the meanwhile I am however due to popular demand consodering adding an optional manual intermediate step to the GUI for the users to to review the pre-segmented text before calling aeneas.

Looking forward to hear your thoughts.

commented

Sure.

While the automated text segmentation works quite well, the descision where to begin a new subtitle is often far away from the segmentation points which subtitlers would choose manually.
One can observe that an automated segmentation prioritizes filling the lines to a maximum (which comes at the price of odd line/subtitle breaks) while manually edited subtitles focus more on a context driven segmentation. This leads to "one line" subtitles once in a while, which doesn't occur in the automated workflow.

The reason why I've asked for the input type "subtitles" is, that I find myself often in the situation that there are already "subtitle ready" segmented texts available. For such a scenario I miss the option to bypass the automated segmentation and do the alignment right away. An intermediate step to have the possibility to edit the auto segmented text before the alignment process would help, however, being able to provide a segmented text to begin with would be preferable.

commented

Ok, cool, closing this issue, unless you have any more questions, ideally I'd like to focus on tweaking the auto pre-align steps, see here

commented

I added an intermedia step in this test release
https://github.com/pietrop/subtitlesComposer-app/releases/tag/1.0.7

see release notes and screenshots for more info

This is in order to do further user testing to improve and refine the pre-segmentation rules by exposing the current output of the segmentation algorithm to the users before sending that off to Aeneas.