Are there people working on an open-source alternative to Aqua Voice (https://news.ycombinator.com/item?id=39828686)
Seems it should be easy enough to have a basic model doing the same, by using Whisper and any sensible LLM
Talk -> streaming STT/Whisper -> some basic processing to put the text in a "markdown-like" format which allows to do structural edits (eg "change the order of the bullet points") -> send to a LLM which process the transcript and apply changes (add text, modify structure, etc)