9 comments

[ 1.3 ms ] story [ 16.7 ms ] thread
I like the concept. If it's truly offline, does it use AI? Do users have to download a local model?
yes it downloads model, and runs it locally.
Working good so far. It's slower than wispr flow
Transcribes great. Two things. - The warmpup takes a couple minutes. I'm guessing it's downloading the local model in the meantime. I'm not sure if there was a reminder in the installer regarding that. - I was kind of expecting it to stream while I talk. Once it works well, I'd set that as default. Nice effort!
You say it's open source, but don't provide any link to the source code at all
I used it but noticed that it puts a lot of load on my memory and it's running a 1 billion parameter model all locally