Your voice might be the most personal data you have. Why do dictation apps ship it to a server?

Here's a test for whatever dictation tool you use: turn off Wi-Fi and try to dictate a sentence. Most of the popular ones go silent, because your voice is being processed on someone else's server.

That always bothered me. Your voice is about as personal as data gets. It carries what you're writing, who you are, and these days it can even be cloned. I didn't want a subscription tool streaming it off my Mac just to type for me.

So I built Rekody: free dictation for the Mac that runs on-device by default. Hold a key, speak, and the text lands at your cursor in any app. Airplane mode included. No account, no audio stored, and no Rekody server your voice could be sent to even if we wanted it. It also has a personal dictionary, so it learns your names and your jargon instead of guessing at them.

The next thing we're building takes "understands you" further: a model trained on your own corrected dictations, so it learns your accent and vocabulary, with your training data deleted after the job. That's the part I want to pressure-test with this community:

- Where's your line on voice data? Is cloud fine if retention is zero, or is on-device the only acceptable answer?

- Would you use a model trained on your own voice? What proof of deletion would you need to trust it?

- What's actually missing from dictation tools that keeps you typing?

We're live on Product Hunt tomorrow. I'll answer everything, including what Rekody doesn't do yet (it's English-focused today).

Brutal honesty welcome.