I’d love to have cmd + k or cmd + l default to voice (instead of keyboard) input that’s transcribed with Whisper. Kinda like on the ChatGPT app, but in cursor. Thanks!
2 Likes
Good idea to be honest
I have created a VSCode extension for exactly this. I just haven’t published it right yet…the project is here though:
This works really well for me and I’ve been testing it for the past few weeks myself so expecting to publish it soon.
1 Like
After seeing this post, I’ve just gone ahead and published the extension 
It’s called Whisper Assistant and Martin Open Sky is the publisher.
Hope it’ll be helpful for you, it has been for me!
2 Likes
No plans to support voice soon unfortunately, but I’m sure other tools will crop up for this usecase.
1 Like
Just posted it on Twitter here:
Awesome work here! If you accept tips, lmk. I’d also love to see support for larger models, or (even better) transcription run off device on OpenAI’s servers (by brining your own key.) In either case, access to a larger model would be nice.
If you have a Twitter/X handle, I can at in you in the post
Hey @aidanmclaughlin,
Glad it’s been helpful to you! Very open to ideas, just getting started really to see if it can be of help in my day to day work.
The only thing to bare in mind is that: the larger the model, the longer the transcription process. I found the small model works well for me with a mic but making this setting easy to change would definitely be helpful! I’m open to ideas so let me know what your experience is and whether there could be more improvements.
I think you found me on Twitter @martin_opensky. I’ve not been so active on Twitter but I’m starting to work with it a bit more recently! Appreciate you tagging me on your post - I’ll likely post any updates to this on my Twitter.
1 Like