Add support for speech recognition

[mrosso10]

Hi, I'm currently using https://github.com/openai/whisper to convert audio to text, and it would be nice to have it integrated to nchat.
I'm currently experimenting with the nchat code and I think I could start a PR with a "minimum viable feature", that would consist of:

• A shortcut that (having an audio message selected) spawns an external (configurable) process
• That process (whisper) would transcribe the audio and save the output to the same folder the audio is in (~/.config/nchat/profiles/...), with .txt extension
• A change to uihistoryview.cpp for rendering the audio transcription near the attachment file name.

Other improvements that would be nice:

• An sign near the attachment filename that indicates if the file is being processed
• Saving the transcription in the chat export

[d99kris]

Hi @mrosso10 - it's a cool idea, but at the same time I'm quite hesitent to adding corner-case features that increase maintainence work (for the core functionality). I suppose we can see how well encapsulated the functionality becomes and then might can consider merging such a PR, but I cannot promise it.

[Bellavene]

Yes, that would be great. While I don't like audio messages at all. Some people send them anyway, and it is a struggle if it has important information, would be great to have a possibility to read it inside nchat.

[mrosso10]

Hi @Bellavene, I'm currently using this patch: mrosso10@7b03de5

It's not a clean code at all because it was just a test for personal use, but it works.

If you could manage to install a speech recognition software (I'm using whisper*)
and want me to help you with that commit, please let me know.

* Warning: whisper eats a lot of memory (~ 4 GB)

[Bellavene]

Hello @mrosso10,
Thank You, I will check it out. Interesting, is there a way to use the system's built-in speech recognition?

[mrosso10]

Do you mean the whatsapp's feature? I don't know, but I doubt it given it's a very new feature.

[Bellavene]

Loosing 4GB of ram for that feature is a bit overkill. I meant the built-in speech recognition in macos.