Italian translation issue - crashes on launch with exception

[nikiromagnoli]

Describe the bug
The program won't start, either CUDA/Normal (I do have NVIDIA card, anyway, and installed CUDA toolkit).
Exception is shown "Failed to execute script 'noScribe' due to unhandled exception: while parsing a block mapping in "" [...]", the whole dialog below.

Screenshots

System (please complete the following information):
Processore Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz (2.21 GHz)
RAM installata 16,0 GB (15,8 GB utilizzabile)
Tipo sistema Sistema operativo a 64 bit, processore basato su x64
Graphic Card: NVIDIA GeForce GTX 1050 Ti

• OS: Windows 11 Pro (25H2)
• On Windows: version? Both. (I tried CUDA version first)
• noScribe Version: 0.6.2
• noScribe Settings: n/a

[kaixxx]

Sorry, there is an error in the Italian translation. I thought I had fixed it, but I may have accidentally reintroduced it somehow.
Anyhow, it is easy to fix:

• Open the folder "C:\Program Files (x86)\noScribe_internal\trans"
• Replace the file "noScribe.it.yml" with this one: noScribe.it.yml

I hope this helps. I'm going to release a new version in a couple of days that has the fix (together with many other improvements).
The fix works for CUDA as well as the normal version. Not sure if the CUDA still works with the GTX 1050 Ti. If you are running into problems, open C:\Users\USERNAME\AppData\Local\noScribe\noScribe, open the file config.yml and change the values for both pyannote_xpu and whisper_xpu to cpu.

[nikiromagnoli]

It worked! Naughty dot ... :)
Now I can give it a try at last! Thank you for your support.

[kaixxx]

OK, I'll leave this issue open until the new version is released, so people can find it.

[nikiromagnoli]

By the way, I gave it a try and it's pure paradise ... I do really appreciate it, thank you.
CUDA version did not work indeed, had to change those two settings to "cpu".

[kaixxx]

Do you remember what error message you got with CUDA enabled?

[nikiromagnoli]

It was sort of a segmentation fault.
Please give me some time to trace it back.

[nikiromagnoli]

My bad, it was a memory overflow:

Conversione audio in corso...
Conversione audio completata

Trascrizione in corso...
Caricamento di whisper

Errore nella fase 3 - trascrizione
CUDA failed with error out of memory

This is it. The progress percentage stalls on 2% of Whisper loading.

[kaixxx]

Ah OK, thank you. You have a 4GB card, this is probably not enough for the large whisper model. You might, however, be able to use the "fast" setting in noScribe, which uses a smaller model with only minimal quality reduction. It depends if you have a lot of transcripts so that the possible speed improvements are really worth the effort.

[nikiromagnoli]

Ok I'll probably give it a try then. Transcriptions I need are not so many; they are long instead (~2h each).