It is time to support speech to text in terminals #1429
Description
The LLMs for speech to text are getting ridiculously good, and at the same time, on Linux at least, getting dedicated dictation apps to work is usually impossible or near impossible. The reason is that Wayland and similar do not have good APIs for this, so it is a train wreck.
OTOH, implementing this directly in the terminal is easy. In the future, there could maybe even be escape sequences for asking the user for input and similar.
So this issue is about rio adding this feature. I know it is not a common feature to have, but the need for this is growing exponentially as people, currently programmers mostly, but soon everybody, start using LLMs to "translate" what they write into some other language, for example through vibe coding.
However, for mere shell work, you can now also do the same, for example uttering: "aichat list the files in the current directory", then the aichat program will be able to do this for you by translating the text into the shell "language". Another data point - it also seems that a significant portion of users of opencode is using voice input.
I am particularly worried about Linux users not having a good setup for this type of work, and I believe integrating this into the terminal could be very beneficial.
There might be some different design choices for how to integrate it, but looking at Handy https://handy.computer/, the interface is quite simple. Basically just choose a model file, a keyboard shortcut, and whether you want push to speak.