Accuracy & Speed

Dictation tools live or die by accuracy and latency. Voice Type improves both by conditioning the audio before recognition and avoiding network bottlenecks.

Faster for long dictations

  • Streaming in 30‑second windows keeps transcription responsive regardless of total recording length.
  • When you stop dictation, only the final ~30s block remains; on an M1 Mac this typically finalizes in ≈2–3 seconds.
  • No uploads means no waiting on coffee‑shop Wi‑Fi or hotel networks, and no compression artifacts.

Higher recognition confidence

  • RNNoise VAD removes silence and suppresses background noise before recognition.
  • Normalization & filtering (−14 LUFS, 50 Hz HPF, band‑limited resampling) help match the input to what Whisper‑style models expect.
  • We avoid heavy “prompted fixes” that can make transcripts look confident but less faithful.

Trust signals

  • App Store sandboxed distribution.
  • Authentic App Store reviews—including critical ones—visible on our listing.
  • Optional BYOK rewrites; otherwise no network calls during dictation.