Latency as the product
Pick it if speed is the thing you notice most and privacy is not a constraint. It feels quicker than the leader for a third less money.
Aqua's bet is that a purpose-built model beats a general one at this specific job, and in English the bet looks right: transcripts land quickly and punctuation is usually correct without a rewrite pass.
The narrower language list and cloud-only architecture are the price of that focus.
Its own, called Avalon, rather than Whisper.
Not currently; it targets Apple platforms.
Which dictation apps handle Spanish properly: inverted punctuation, regional accents, tuteo vs usted register, and Spanish-English code-switching, tested across the full ranking.
Inside the cleanup pass that defines 2026-era dictation: filler removal, punctuation, formatting and tone adjustment by a language model after transcription — why it feels like magic and where it goes wrong.
The 2026 engine landscape: Whisper's 99-language legacy, Parakeet V3's European sprint, and the proprietary cloud models pulling ahead at the top end — plus how to find out which engine your app actually uses.
What dictation apps actually do with your audio and transcripts: cloud retention policies, training opt-outs, what 'on-device' does and doesn't guarantee, and how to choose for clinical, legal or confidential work.
The state of voice coding in 2026: why identifiers and symbols break general dictation apps, how command-mode and context-aware apps differ, and practical setups for wrists that need a break.