Subtitles with long lines tend to cause GPT desyncs, as well as producing translations that are too long. To avoid these problems it would be helpful to pre-process the source file looking for long lines (by duration?) and split them at newlines, periods or commas.
Splitting the duration proportional to the length of the subtitle tends to produce pretty good results in Subtitle Edit, subject to a minimum duration. This functionality could be added to NewProjectSettings prior to the batch preview (maybe even incorporated into SubtitleBatcher), so that the resulting batches would still respect min/max line counts.
machinewrapped
commented
5 months ago
Subtitles with long lines tend to cause GPT desyncs, as well as producing translations that are too long. To avoid these problems it would be helpful to pre-process the source file looking for long lines (by duration?) and split them at newlines, periods or commas.
Splitting the duration proportional to the length of the subtitle tends to produce pretty good results in Subtitle Edit, subject to a minimum duration. This functionality could be added to NewProjectSettings prior to the batch preview (maybe even incorporated into SubtitleBatcher), so that the resulting batches would still respect min/max line counts.