Voice to text?

[pierceglenn12345]

Are there any plans to insert an initial step of using something like vosk, pocketsphinx, spchcat, or whisper?

One one hand spchcat would be nice for its local first approach, where as on the other end whisper would require another subscription with openai.

Personally I like the idea of verbally communicating with a "personal assistant" for org-mode purposes.

Just gave this a try for the first time today and it is truly an incredible feat! It was only a matter of time before these text processing ai's started directly communicating with the one true text editor. :)

[belden]

@pierceglenn12345 I'm on Ubuntu 24.04 and run emacs from snap.

About a month ago I went through and set up https://github.com/PhilippeRo/IBus-Speech-To-Text and taught emacs to pay attention to ibus input by adding some exports to my ~/.profile -

export GTK_IM_MODULE=ibus
export XMODIFIERS=@im=ibus
export QT_IM_MODULE=ibus

If you're also on Linux then I'd suggest having a look at ibus-speech-to-text, which uses vosk models under the hood.

I can report that speech-to-text works well enough; it's not perfect, but it's sometimes my preferred input mode.

I think that speech-to-text is an independent concern from efrit. Perhaps this should be closed?