Voice poor performance

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I’m not sure who has been tweaking the voice dictation for cursor, but this is a no-go. Why mess with something that worked so well. Fix it or I have to start looking for another IDE.

Steps to Reproduce

Just try and use the voice dictation… waist of time at this point.

Expected Behavior

It works as before (last week?)

Operating System

Windows 10/11

Version Information

Version: 3.3.22 (user setup)
VSCode Version: 1.105.1
Commit: 38a27120cfc7419a5efa38420665eaeeed1e7b30
Date: 2026-05-07T07:47:13.552Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Windows_NT x64 10.0.26200

Does this stop you from using Cursor

No - Cursor works, but with this issue

7

We’re aware of voice dictation quality issues that appeared in recent builds. Our team is actively investigating.

A few things that would help us match your experience to the specific bug being tracked:

  1. What specifically changed? For example: is the transcription producing garbled or duplicated text, is it slower than before, or is accuracy worse?

  2. Where are you dictating? The editor (mic button in the editor area) vs. Composer/Chat can behave differently.

You may also find this related thread useful: Voice input — silent stream / transcription issues

8

So my mic is quite functional… the only thing not functional is cursors chat… its horrible now.. it does not conform to the chat composer, its a single line, if I press send before some magical time frame then it truncates my input, it starts to just copy a sentence in run on for tens of lines… guys this was working beautifully for months… I’m assuming you are trying to save a buck with a less capable compute… not cool, the app is not useable in this state…

10

I vibed this since it’s so broken, local and on your machine, uses a bigger whisper model also so the results are better GitHub - jon-makinen/genie: Local open source Whisper dictation as a macOS menu bar app in any input field · GitHub

12

I have the exact same issue on Mac. Previously it was one of the best voice-to-text experiences I’ve had, but in recent updates I have to fight with it. It either completely mis-transcribes what I’ve spoken, or in other cases you speak and only the end of your speech shows as text, and after a short delay the rest of it comes or might not! which is really confusing and makes it slow.

The other annoying thing is: why does the voice button disappear when I put an image or text in the input field? This makes me paste the text to some markdown file just to get the voice button back, so I can add what I want, then I have to paste my previous text back in before it!

Why can’t we have the voice button always visible? So when you already have some text or images in the input field, you can turn it on again to add more if needed.​​​​​​​​​​​​​​​​

13

Thanks for the detailed description. The transcription quality issues you’re seeing (mis-transcription, partial/delayed text) are being actively investigated. Your report is helpful because it confirms this affects Mac as well, not just Windows.

Regarding the voice button disappearing when you have text or images in the input field – that’s a separate known issue. We have an open request to keep the mic button visible alongside the submit button at all times, which would solve the workaround you’re describing.