What would you like to see?

Currently once we end STT session we immediately send the transcript as a prompt to the LLM provider. Sometimes its annoying when -

1. STT recognises some words incorrectly but still send it as a prompt before giving me a chance to correct it.
2. I need to think while speaking but if the pause is too long (currently more than 3.2s) it submits the incomplete prompt.

One way to achieve both of the above issues is if we can put the recognised text in prompt input area (already happening) but not send it till user clicks on send button or some equivalent action (min one extra click required from user). This can -

1. give me a chance to correct the incorrectly recognised words.
2. let me add to existing text (in between or at the end)

As far as I have searched, I found no existing feature request similar to this hence opening a new issue.