I was casually scrolling HN yesterday when I came across the post comparing whisper small with Apple's new native SpeechAnalyser API, where it beat it in both accuracy and speed [1] so I built it into a fully native, OSS dictation swift app that does everything I need.

Megaphone is built on top of Freeflow and uses the SpeechAnalyser API along with Apple's own foundational model for intelligence - both running on the NPU so it's efficient, it also streams the audio while you speak so it's done processing ~<1s after you're done speaking.

Usable for all Apple Silicon macs running MacOS 26 (Tahoe) or higher.

[1] - https://news.ycombinator.com/item?id=48894752