v0.73.0 · macOS 14+ · Weekly ships

Speak your mind,
captured forever.

A macOS menu-bar voice input. Hold the hotkey, speak — text streams into your cursor in real time, and every sentence sediments into a searchable, private memory layer that lives only on your Mac.

Download v0.73.0 See how it works →
macOS 14.0+· Apple Silicon · Intel· 22.6 MB· Completely free
THREE WAYS TO USE IT

Free forever, two ways. Pro when you want zero setup.

Local engines and Bring-Your-Own-Key paths are free forever, no subscription. Pro removes all cloud quotas with zero configuration.

Free · Local

100% Local

$0forever

Three on-device ASR engines (SenseVoice / Paraformer / Apple). Nothing leaves your Mac.

  • ✓ Open the app and dictate
  • ✓ Zero network needed
  • ✓ Local-only AI tidy fallback
Free · BYOK

Bring Your Own Key

$0forever · pay your provider directly

Plug in any OpenAI-compatible API key (DeepSeek / Kimi / OpenAI / your local OpenAI-compatible server). Unlimited tidy, you control the cost.

  • ✓ Unlimited AI tidy
  • ✓ Pay only your token cost (~$0–2/mo)
  • ✓ Works with any OpenAI-compatible endpoint
  • Will always be free — never gated behind Pro
Pro

VoiceInput Cloud

$9/mo · or $79/yr · or $49 lifetime

Zero-config cloud ASR + cloud AI tidy. Removes the 60 min/mo + 50 tidy/day quota. For people who just want it to work.

  • ✓ Unlimited cloud ASR
  • ✓ Unlimited cloud AI tidy
  • ✓ Up to 3 devices per license
See pricing →
PRODUCT · THREE LAYERS

One thing. Three depths.

Input on top, voice archive in the middle, AI memory underneath.

L1 / TOOL
Tool SPEAK

Hold, speak, release. Mixed-language, homophones, fillers handled silently. Under 1.4s.

L2 / DATA
Data RECALL

Every line archives locally with source app, time, tags. Search, filter, export.

L3 / MEMORY
Memory REFLECT

7 personas review your week. A weekly MBTI sketch. 3–5 quotes worth echoing.

AI MEMORY

Same line. Different reader.

7 built-in personas plus your own. Contrast itself is the memory.

Entry · 2026-04-18 14:02 Work / Product decision
💼Boss 🎯Coach 🚀Musk 💡Jobs 🧘Therapist 🤝Friend ✍️Editor
"The current design feels pointless, but the team spent three weeks on it — cutting is expensive…"
🎯 Coach Sunk cost isn't a reason to continue. Real question: is continuing lower ROI than alternatives?
↳ If those three weeks never happened, would you pick this today?
This week · Apr 13 → Apr 18 Big5-based
INTJ
I
E
N
S
T
F
P
J

Introspection up, decisions slower. Three returns to the "sunk cost" theme.

"Speed is iron law — no feature may add perceived latency."

04-17 · Xcode · ★
↳ When speed and correctness collide, which yields first?
ENGLISH POLISH

Output that reads like writing.

Casing, punctuation, and unit spacing — handled locally in <5ms, no LLM call.

Live preview · typical cases 4 rules
S1
hello,world
helloworld
S1
use kimi to design an api
use Kimi to design an API
S2
(about deepseek api)
(about DeepSeek API)
S2
deepseek responds in 50ms
DeepSeek responds in 50 ms

LLMs handle semantic judgment (homophones, fillers). The local engine handles format (brand casing, punctuation spacing, unit spacing). Two layers; both wins.

45+ brand names auto-corrected. Spacing after English commas/periods. Half-width parentheses kept for ASCII. Every rule toggleable.

<5ms
all rules run
0
network calls
Aa
auto brand casing
50 ms
unit spacing
DETAILS

Every detail has a reason.

Target window lock

Pin source app at record-start. Switch windows — still lands right, with clipboard fallback.

v0.12.0
Invisible cleanup

Stop talking, wait 1–2 seconds, polished text lands in one shot. You never see the raw.

v0.12.0
Pinyin disambiguation

Local pinyin injected into prompt. Homophone pairs no longer confused.

PY-GEC
Learning loop

Your edits on AI output extract as candidate rules. Accept from menu bar.

INBOX
200+ default hotwords

AI models, dev tools, Apple products built-in. "Cursor" stays "Cursor".

CODE-SWITCHING
Adaptive overlay

More text, more transparent. Breathing glow hints AI is working. Original never disappears.

5 SIZES
PRIVACY

Data stays local. Promises stay explicit.

Audio, text and history all live on your Mac. One single number leaves — seconds per recording.

LOCAL

Everything stays on your Mac

Audio and text land in the app's own directory, auto-backed up on launch. Uninstall takes it all with you.

PULSE

Only one number goes out

Each recording sends only its length to the global pulse. No identity, IP, content, or context. Toggle off in Settings.

KEYS

Keys live in Keychain

API keys sit in macOS Keychain, never on our servers. ASR runs directly against Volcengine, nothing persisted.

CHANGELOG

Last seven releases.

We don't pile features — we only ship what we believe earns its place.

v0.73.x 2026-06-01

Faster + more accurate translation

Cloud recognition / AI cleanup is faster: connection reuse makes text appear sooner after release in most cases. Translation / bilingual is more accurate: proper nouns (company / product names) are recognized better.

v0.72.x 2026-05-31

New AI Translation + AI cleanup answer-bug fix

New AI Translation: speak and get the translation directly, or original + translation side by side, 50+ languages — switch "Tidy / Translate / Bilingual" from the menu bar. AI cleanup is more accurate: fixed the occasional case where it answered your question instead of just tidying what you said.

v0.71.x 2026-05-26

Cloud speech recognition starts up faster

Cloud speech recognition starts up faster — pressing right Option begins recognition almost instantly. Fixed occasional response stalls.

v0.70.x 2026-05-24

AI cleanup back to generation-leap fast + onboarding upgrade + 3-tier auto-update prompt

AI cleanup speed massively improved — release-to-text feels generation-leap fast again. New-version updates are now harder to miss: after 24h the banner turns red, after 48h it auto-restarts to finish the update. Onboarding redesigned with a full-size keyboard and three-phase animation (press → speak → release & text appears) so first-time users know which key at first glance. Dashboard adds a one-click fix entry when Accessibility permission becomes stale.

v0.69.x 2026-05-20

AI cleanup noticeably faster + zero-setup + all-new onboarding

AI cleanup is noticeably faster — release the key and polished text appears almost instantly, no configuration needed. An all-new onboarding flow helps first-time users get up and running right away. Long-sentence direct insert is more reliable and produces more complete results. Recording status is clearer and more polished, proper-noun recognition is sharper, and sign-in on certain networks is fixed.

v0.68.x 2026-05-15

Faster, more accurate AI tidy + auto-fix for misheard names & idioms + chat / email tone preserved

AI tidy response is markedly faster — the release-and-see-text loop feels much snappier. Common AI company / tool names (Anthropic / OpenAI / DeepSeek / Cursor / TypeLess) now hit correct spelling far more reliably. Chinese idioms that get misheard (e.g. 头痛医头脚痛医脚 / 实事求是) are auto-restored. Chat & email contexts preserve your spoken tone — no longer forced into stiff business prose.

v0.66.x 2026-05-13

Local-insert duplication fix + AI tidy more reliable + learning loop

Fixed local-insert occasional text duplication (when ASR re-identified a sentence mid-utterance, the cursor showed duplicated content). AI tidy more accurate: company names / tech terms / homophone corrections improved; output stays closer to your original phrasing. New learning loop: after you manually correct an AI-tidy output, the system remembers the proper-noun mapping and gets it right next time automatically.

FAQ

Things you might ask.

Do I bring my own API key? Cost?+

Yes. ASR is Volcengine, LLM can be Doubao / DeepSeek / Kimi / OpenAI. Full control over account and bill. Typical: CNY 5–20/month.

How does it compare to TypeLess / Wispr Flow?+

On Chinese scenarios, much faster end-to-end (1.4s vs 3–10s). And it's not just input — everything you say becomes a searchable memory.

Systems? Intel Mac?+

macOS 14.0+, Apple Silicon + Intel. 22.6 MB DMG, non-App-Store, Sparkle auto-update.

Permissions?+

Microphone, Input Monitoring, Accessibility. Granted once via the onboarding page.

Will AI tidy mangle what I meant?+

No. Prompt constrains LLM to three jobs: fix homophones, drop fillers, add punctuation. Confidence < 0.5 keeps the original. Double-tap right Option to bypass AI.

Export and migrate?+

Yes. Markdown / JSON / CSV export. Copy the DB file to the same path on a new Mac.

Turn off the AI memory layer?+

Yes. Clear API config, all memory features stop. Local typography engine keeps running.

COMPARE

Still picking a voice input app?

Honest, side-by-side comparisons against the tools you're probably also evaluating.

VoiceInput vs Superwhisper

Faster on Chinese, plus a memory layer Superwhisper doesn't have. Read the side-by-side →

VoiceInput vs Wispr Flow

Different categories: Wispr rewrites tone, VoiceInput keeps memory. Compare →

VoiceInput vs Apple Dictation

No 60-second cap, AI cleanup, mixed-language handling, full archive. See full →

Every word matters.

Download · grant three permissions · hold right Option. Thirty seconds to get it.

Download VoiceInput_v0.73.0.dmg · 22.6 MB
macOS 14.0+ · All versions