Whisper is pretty good for speech-to-text.

Groq Whisper is very fast and cheap. I think up to 8 hours per day is still free through their API?

aider has a /voice command.

voice.sh, a fork of aider's voice command made for the terminal.

youtube-text.sh

.vimrc configuration for voice command:

nnoremap glv :r !voice.sh 2>&1 \|
            \ tee /dev/tty \|
            \ sed '/^Recording, press ENTER to send to Groq when done/d' \|
            \ sed 's/<span class="createlink"><a href="https://secure.diyhpl.us/write/diyhpluswiki/ikiwiki.cgi?do=create&amp;from=voice&amp;page=%3Aspace%3A" rel="nofollow">?</a>:space:</span>//'<CR>

Summon a terminal and just speak