I’m holding down the F1 key while I speak using Wispr Flow. I’m not pressing anything on the keyboard, so this is a genuine test, and I’m speaking at full conversational speed right now. I tried using a Shox head-worn mic and bone-induction headset, but it was really uncomfortable and I couldn’t lie down in bed with it because it pressed against my pillow and flopped around. The audio was terrible. That headset had about a 10-hour battery, which was okay. But it was just annoying to have it on my head all the time. Then I realized there was something called a DJI Mic Mini. So I got one of those for about $130 from Amazon. And then I began to experiment with how to work around the terrible cursor microphone button.
I tried the Windows key plus the H key on Windows to try to get some kind of dictation tool working, but it just kept slipping off the screen every time I used it. Then I discovered something called Voice Access, and after a week of talking to it, it was frustrating as hell because it would only understand 50% of what I said. Then I researched OpenAI Whisper plugins for Cursor, which require a ton of coding and a subscription to OpenAI, which I have. It has no graphical interface whatsoever and is basically a box of nails. Also, after a long period of time trying to link it in with an extension, it became clear that Cursor was just broken, or it had made modifications without any warnings and left dead code littered throughout its extension library.
It confused me to discover there was something called Wispr Flow. A similar name, but I’m not sure that it actually uses the OpenAI system at all. This one is free for a month, and then you pay about $12 a month after that, apparently. For me, this is the first hour of using it, and I’m extremely happy at its response. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10. I just wanted to see if I could do that.New paragraph.New paragraph.
I guess it doesn’t understand new paragraphs. It thinks I want to write a new paragraph. I guess I could switch some sort of mode in the settings to make that happen, but it’s probably a good thing that that doesn’t override normal dictation. As you can see, the punctuation is good, and hopefully it understands individual words like “in,” “of,” “at,” “the,” “so,” and “to.” Wow, it even put inverted commas around it.I slurred that last sentence, and it understood me anyway. There’s also a loud fan blasting me from the side because it’s a hot day.
It would have taken me a long time to type all this, but I’m just blurting out whatever I want to say. Let me just check some coding words: component, element, header, footer, panel, admin-, new.js. Show status (open) function. I guess it doesn’t know exactly what I want, but at least it doesn’t misinterpret me. I’ll try just saying file extensions. JS, CSS, HTML, exewow, it knew how to separate all those letters.I guess it didn’t know how to separate EXE from WoW.Even so, it’s much better than Windows 11 voice access, which is absolutely terrible.And it’s definitely better than me having to sit and type all day.I highly recommend this combination.
-
DJI Mic Mini. You get two transmitters, so you get 10 hours per transmitter, and one is always charging, so you get infinite microphone wireless that you can walk around the house with.
-
Wispr Flow. It does require you to hold the button on the keyboard while you’re talking and then let go when you’re finished.
Wow, it even made it into dot points. That’s pretty cool. It kind of knew what I was doing.From context.The only thing it doesn’t seem to be able to do is put a space after a full stop.Even so, it’s not bad. And it will just type in whatever text input is being focused on.So right now, it’s typing into this box on the Cursor Forum, but if I click over there, on the right, where my cursor interface is, it will type what I say in there instead.
