[_mrinalwadhwa_]

I think realtime transcription hurts the UX of polishing what's said worse. In FreeFlow the output of the transcription is fed to an LLM to polish in context of where the text is being injected. This way we can go beyond naive transcription.

FreeFlow already feels extremely fast and text being typed as I dictate is distracting especially if the polishing phase edits it.

[Leftium]

I would delay polishing until right before delivery.

Eventually, I will add a polishing step to my own https://rift-transcription.vercel.app.

Right now, you can experience what true realtime streaming transcription feels like.

I plan to add two "levels" of polishing:

- Simple deterministic text replacements will be applied to both interim and final text.

- LLM polishing will only be applied right before delivery.

- It will be possible to undo one or both polishing steps. (Actually even more fine-grained undo: at the replacement rule level).

[_mrinalwadhwa_]

That said, FreeFlow is open source for exactly this reason, everyone will have their own preference. If you would like to turn this behavior into a configurable preference, we'd happily accept a pull request.