First of all, I really love the Cursor IDE — it has genuinely changed the way I work as a developer. Thank you for building this product.
After using it for a while, I noticed that typing long prompts in the chat takes a lot of time. I’d love to use voice input instead of typing to speed up my workflow. However, the current voice input feature has very high latency and feels laggy — after I finish speaking a sentence, it takes a long time before the text appears in the input box. The experience is far from smooth.
Honestly, I wish I could just “talk to Cursor” in real time to save the communication cost between me and the AI.
My request:
Please consider optimizing the voice input experience. Ideally, I’d love to see real-time (streaming) speech-to-text, similar to what modern dictation tools offer — text appearing while I’m speaking, not several seconds after I stop.
This would make Cursor even more powerful for developers who rely heavily on the chat to collaborate with the AI.
Really looking forward to this. Thanks again for changing the way I develop!
Thanks! I was asking because the Agents window and the editor actually use two different voice input implementations (he editor streams text as you speak, which sounds closer to what you’re looking for, while the Agents window waits until you stop speaking to transcribe the full clip). So you may already have the better experience of the two.
No specific improvements are planned right now, but we’d love to hear from other voice mode users as well!
@Colin any plan of switching the Editor window to the Agents behavior so it also waits until I stop speaking? it is much much more accurate.. or what is the reason behind keeping these two split?
@liquefy I think this thread demonstrates that different users have different preferences, and we didn’t want to disrupt the existing workflow in the Editor.