Subtitle Edit 4.0.1

[Old man Alan]

I am an aged user and I find Subtitle Edit 4.0.1 a fantastic program.
However, when auto transcribing subs from the audio it continually and emphatically writes [MUSIK] in the titles even when i can't hear any - is there a way of removing the text [MUSIC] without going through each and every individual subtitle.
Possibly something like “Find and Replace” or some where in settings to stop it writing it in the first place.

[Ennio]

There's two ways of calling the "Replace" function. One is to click on "Edit" in the upper menu bar. This opens a list where you'll find it.
The other way is to press and hold the "Ctrl" key on the keyboard and type H.

In the "Find what" box type [MUSIC] (with or without space before/after)
Leave the "Replace with" box empty and click on "Replace all"

Removing unneeded spaces can also be done with "Fix common errors". You'll find it in the "Tools" section (next to Edit). There's also a button to call it; it's the two crossed handtools.
In the list of the tool there's an option to check "Remove unneeded spaces".


Hope this helps

[Old man Alan]

Thank you Ennio that's brill

[Ennio]

You're welcome, but I failed to mention the most obvious way since MUSIC is written between [ ]. These brackets are commonly used for descriptions for the hearing impaired. SE has a tool to "Remove text for hearing impaired". It's also in the Tools list. The button for it is the big ear. Or press and hold "Ctrl" and "Shift" keys and type H. The tool will auto-remove left-over spaces as well, btw.

[Old man Alan]

I find that when using “Audio to text (Whisper)…” there are some quite large portions where there isn’t any subs just Start time - End time and a blank space in the Text box is there a way of fixing this?
It’s as if “Audio to text (Whisper)…” simply doesn’t translate these portions.

[Ennio]

In the aforementioned "Fix common errors" tool, there's an option to check "Remove empty lines/unused line breaks".
As for "Audio to text (Whisper)", I have never used it so can't really recommend on how to possibly avoid the issue you mentioned.

[Old man Alan]

Thank you Ennio you have been extremely helpful

[Subtitles]

I find that when using “Audio to text (Whisper)…” there are some quite large portions where there isn’t any subs just Start time - End time and a blank space in the Text box is there a way of fixing this?
It’s as if “Audio to text (Whisper)…” simply doesn’t translate these portions.

Which Whisper Engine have you selected (default engine is cpp but there are 5 more some need installation on your PC in order to be able to use them), and which transcription model are you using? (choose model medium, base, small, etc).
If you don't have a GPU installed then you might get issues trascribing using the more accurate models like medium or large.
What is your CPU model?

[Old man Alan]

Here are my Whisper details and GPU

Engine = Const-me
Model = base (141 MB)
GPU = NVIDIA GeForce GTX 1050 Ti

[Subtitles]

Here are my Whisper details and GPU

Engine = Const-me
Model = base (141 MB)
GPU = NVIDIA GeForce GTX 1050 Ti

Thanks. Your GPU has 4GB VRAM which is good.
Can you upload only the audio file (DM please if you prefer) and I will try to replicate the issue.
I usually transcribe the audio instead of translate, unless it is in English.
If your audio is in English, then use model base.en

[VoodooFX]

Engine = Const-me
Model = base (141 MB)

Use better engine and use larger model.

[Old man Alan]

I should have said Transcribe instead of Translate sorry.
I’ve tried it again using “Audio to text Whisper” with settings changed to;
Engine = CPP
Model = base en
and it is much better.
Thankyou

[VoodooFX]

Use not smaller than medium model, and faster-whisper engine.

[Old man Alan]

How do i know what is a big or small engine they just have names?

[VoodooFX]

Model.

[Old man Alan]

I called myself Old Man Alan because i am getting on to 80 years old and i don't understand you just saying Model.
In models I only have two choices “base” and “base en” how do I know which of those is big or small?

[VoodooFX]

Then download bigger models.