At the moment, when using Voice Mode, you can record your voice only once - after transcription, the microphone button disappears and is replaced by the “Send” button.
Because of that, it’s impossible to either add more voice input or re-record without first clearing the entire input field.
It would be very helpful if the microphone button stayed available after the first recording, allowing the user to:
add more text by voice.
This would make Voice Mode much more practical for longer prompts or when refining thoughts in multiple short recordings.
I’ve also noticed if you add tabs or any other context before adding your own prompt, voice mode disappears because the submit button replaces voice input, now that the prompt input is populated. Voice input should be a completely separate button that has no contextual changes, so it’s always available to use anytime.
Please please implement this. Otherwise I need to pay for yet another subscription just to do speech to text when instructing my agents. It can just be deducted from my included usage.
While waiting for the continuation mic feature to be added, as a work around you can open a new chat tab and speak there. Then copy paste into original chat.
Would be nice to get an improved STT though. Something like ElevenLabs Scribe v2 Realtime. I think currently Cursor uses Browser’s free speech recognition which isn’t that accurate. I wouldn’t mind if this was consuming Cursor usage. It’s just so convenient to use voice, if I don’t want to bother using keyboard.