sometimes create malformed SRT files that don’t work

[ClaireCJS]

Now that I’m using this as part of my whole-collection AI transcription system .... I’ve definitely exposed an outright bug where SRT files are so malformed that the timestamps are displayong on WinAmp’s MiniLyrics plugin [which i haven’t seen in 5+ years of using it].

I’m getting output like this:

18
00:00:34,470 --> 00:00:36,570
the

19
00:00:36,570 --> 00:00:38,890
20
00:00:39,990 --> 00:00:42,350
21
00:00:42,350 --> 00:00:44,910
22
00:00:47,330 --> 00:00:48,110
23
00:00:48,110 --> 00:00:48,470
tower

Those blank lines shouldn’t exist. Or at least should have a space or something so that it’s a null timestamp.

Most parsers are going to run into:

19
00:00:36,570 --> 00:00:38,890
20

and simply display "20" as if "20" was what was meant to be displayed.

But no.
WhisperTimeSync is forgetting the 3ʳᵈ line, and the line in between, in certain cases.

Also putting orphan lyrics up top with no timestamp so they never display

Attached are the files for this situation

nomeansno-thetower-official-lyrics.txt

nomeansno-thetower-whisperai-output-srt.txt

nomeansno-thetower-WhisperTimeSync-malformed-srt.txt

[EtienneAb3d]

Hi Claire,
WhisperTimeSync is trying to put the good time-stamps over the good text.
In your case, you have too much time-stamps on the input SRT because "the tower" is repeated, while it appears only once in the input TXT.
In this case, the real question is: what file is wrong? Looking at the duration, I guess the lyrics TXT is not provided with the repetition that is really in the song.

[ClaireCJS]

Yup! Human nature is to not ever do lyrics correctly with repeats. I hate it.