For people who
live on the keyboard,
a voice that types
Hold a key. Talk. Words land where your cursor is.
100% on-device · CJK fluent · Native on Apple Silicon.
macOS 13+ · Apple Silicon · 0.3.0 beta
Hold a key. Talk. Words land where your cursor is.
WisPod is dictation done so quietly you forget it's there. It doesn't write for you — it just turns what you say into the text wherever you're typing.
On-device inference
Audio never leaves your Mac. Models run natively on Apple Silicon, working perfectly even with Wi-Fi off.
Words appear as you speak
No waiting for the end of a sentence. Every pause is the product's working tempo.
CJK without switching
Mix Chinese, Japanese, Korean and English in one breath. Punctuation, numbers, and terms tidy themselves up. Tone shifts cleanly between code, docs, and chat.
Works inside the apps you already use.
Not yet another transcription window. Wherever your cursor is, that's where the words land.
$ git commit -m "refactor: swap fetchUserProfile for a useUser hook with a 5-minute staleTime"
[main 7f3a2c1] refactor: swap fetchUserProfile for a useUser hook…
1 file changed, 6 insertions(+), 14 deletions(-)Wherever your cursor goes, WisPod follows.
Anywhere on macOS that takes typed text. Code, docs, messages, notes, email — talk, and the words land at the cursor.
Three local tiers. Pick yours.
All three run on-device. They differ only in size and personality. Short bursts → Light. Daily use → Recommended. Tackling jargon → Heavy.
All three run locally. Models stay on your machine. The cloud never sees them.
100% local.
- →Audio never uploads. Not one byte leaves this Mac.
- →Model weights live locally. Inference runs on Apple Silicon.
- →Works offline — on a plane, behind a firewall, anywhere.
- →No accounts. No tokens. No cookies.
Chinese, English, Japanese, Korean — all in one breath.
No language picker. No mode switching. WisPod handles the seams between languages itself.