The following was dictated without a single correction, using Whispering v7.9:

I am yelling at my computer from on an exercise bike to dictate this blog post and tell you that at long last there exists open source dictation software for Linux that actually works. I am using Whispering developed by Epicenter to dictate this blog post so you can get a sense of how accurate the transcription is without editing. I'm sure there will be mistakes but it's a lot better than anything else I've tried. Under the hood it can use a number of providers for transcription including Speeches which is open source but I'm using the simplest that is available for offline transcription. Whisper C++ and it seems to work relatively well. I have used Whisper C++ on its own for transcribing recordings with good results but now there is a GUI app that makes the recording and transcription easy for anyone even people who aren't technical. Not all computers I have tried to install whispering on have got along with that but if you are able to install it and it works it's the best open source dictation software I've found.

Note: I used Whisper C++ with the Medium model, using CPAL recording method, pipewire as the device, the recommended compression settings, and 16 kHz as the sample rate. I didn’t notice any improvement when I re-transcribed using the Large v3 Turbo model.