YapVault, the voice recorder that knows who was talking

Field notes

The proposed workflow combines speaker separation, on-device transcription, searchable transcripts, and playback links that jump to matching words. The page gives a concrete storage comparison—about 14 MB per hour versus 29 MB for a typical recorder—and says the first speech-model download is required only once. Apple Intelligence is required for automatic titles and summaries, while the page says core recording and search work offline afterward.