MicType

Dictation is now faster — and you can watch it happen. · Speed: cloud dictation typically lands in about a second. Recordings upload ~5x smaller, the cloud races your on-device engine with a hard time budget, and a slow network can never stall a result. · Accuracy: cloud recognition now runs on a higher-accuracy engine, your words stay in the language you spoke (mixed Chinese-English included), and the polish step is smarter about fixing misheard words. · Pick your transcription service: Cloud recognition (optional, off by default) now has its own provider choice — with a one-tap connection check that shows the round-trip time. Use a fast transcription provider while your AI assistant stays wherever you like; picking the same provider for both shares a single key. · Live draft: while the final text is being prepared, the keyboard shows a clearly-labeled draft inside its own panel. Your message field only ever receives the final text, once. · Custom reply (new): save a reply style in Settings (e.g. "answer with dry humor"), then copy a message and tap the wand key — a reply in your style is drafted by your own AI and inserted at the cursor. · Polish keeps your voice: fillers, stutters and misheard words get cleaned up, but your tone and phrasing stay yours — chat messages still sound like you, and every fact, number and question is kept. · Reliability: a result can no longer appear in the wrong app long after a dictation, consumed text never lingers, and empty or silent recordings say "didn't catch that" instead of inserting junk. Thanks for using MicType!