Frequently Asked Questions

Kaikou is a native macOS menu-bar voice input app. It lives at the top-right of your menu bar: a hotkey starts recording, your speech becomes text, and it inserts automatically at the cursor. What sets it apart are AI polishing and offline speech recognition, giving voice input far better accuracy and professionalism than plain speech-to-text. There is also a Windows edition (direct download from this site — the app is free, AI uses your own key).

Search for "Kaikou" on the Mac App Store and install. The app then appears in the top-right of your menu bar (the K icon). On first launch, grant microphone permission, pick a hotkey mode, and you're ready.

Kaikou is a menu-bar app (no Dock icon); the K icon sits at the top-right of your screen. If it's missing, other apps may have pushed it out of view. Click the "≫" arrows on the right of the menu bar to reveal hidden icons, or drag the Kaikou icon closer to the front.

Kaikou supports 4 hotkey modes:

• Double-tap Command (default) — tap ⌘ twice quickly to start recording
• Hold Right Option — hold to speak, release to stop
• Double-tap Right Command — tap the right ⌘ twice quickly
• Hold Fn — hold to speak, release to stop

Switch anytime under Menu-bar K icon → Settings → Hotkeys.

Text is inserted automatically at the cursor in the active input field. Kaikou pastes via a simulated Cmd+V, then restores your original clipboard contents.

Three steps:

• Install: open the App Store on their Mac, search "Kaikou", and install (core features are free)
• Authorize: grant microphone permission on first launch and follow the guide
• Use: open any chat or text field, double-tap ⌘, speak, pick a candidate — the text is sent

Walk them through it once; after that they can operate it themselves — no typing needed at any point.

Kaikou was designed exactly for this: from starting recording to text on screen, no keyboard required.

• The default "double-tap Command" needs only two light taps; if even that's hard, switch to hold-Fn mode (release to stop) in Settings
• Text inserts directly at the cursor — no manual copy-paste
• Without the Accessibility permission, text goes to the clipboard with a paste prompt — the app still works

No. Direct text insertion (the core feature) is free forever for everyone, and the free tier includes 100 AI-enhanced recognition credits per day. A subscription is only needed for unlimited AI polishing and terminology-aware translation — everyday use costs nothing.

Worth a try. The offline engine (SenseVoice) tolerates some accents in Chinese — slow down and speak one sentence at a time. If recognition slips, enable AI polishing to fix errors (there's a free daily quota), or switch to the Apple Speech engine and compare results.

Just re-download Kaikou from the App Store on the new Mac and sign in with the same account (Sign in with Apple or phone number). Subscriptions are tied to your Apple ID: sign in with the same App Store account and it's restored — no double billing.

Apple Speech is macOS's built-in speech framework — free and ready with no extra downloads; when online it may use Apple servers for better accuracy.

SenseVoice is an offline engine built on sherpa-onnx — an 894MB model download, recognition runs entirely on-device with no network. It's more accurate for Chinese, especially technical terms, and is an optional add-on.

Try these to improve accuracy:

• Switch to the SenseVoice offline engine (more precise for Chinese)
• Make sure the microphone picks you up clearly; reduce background noise
• Speak at a moderate pace, enunciate clearly
• Enable AI polishing to correct recognition errors automatically

The SenseVoice model is about 894MB. Menu-bar K icon → Settings → Recognition Model → choose SenseVoice, and the app will guide you through the download. The model lives inside the app sandbox and is removed automatically if you uninstall.

Apple Speech supports many languages (per your system's list), selectable in Settings. SenseVoice is currently optimized for Chinese. AI polishing supports the languages your chosen provider supports.

AI polishing sends your raw dictated text to an AI model, which refines it for the current scenario (Prompt / Code / Shell / General). For example, in Prompt mode the casual "help me draw a nice gradient background" becomes a professional-grade prompt. You need to enable AI in Settings and configure a provider first.

Configure it under Menu-bar K icon → Settings → AI Polishing. Kaikou speaks the OpenAI-compatible API and ships presets for 10 providers (DeepSeek, Qwen, Zhipu GLM, Doubao, MiniMax, Kimi, ERNIE, Hunyuan, SiliconFlow, OpenRouter) — just paste your API key. GLM, ERNIE and OpenRouter are marked as "free channels" with links to sign up for free keys. Custom providers work too (enter a Base URL and Model ID).

When enabled, finishing dictation pops a candidate panel with 4 AI-generated interpretations. Say "write a sort function" and candidates might include "implement sorting", "create a sorting algorithm", "write sorting code", and more. Pick with ↑↓, confirm with Enter, dismiss with ESC. Enable it manually in Settings.

AI polishing and candidates send your dictated text to the provider you configured yourself — the traffic never passes through our servers, and we store neither requests nor responses. Choose a provider you trust and read their privacy policy to understand how they handle data.

Three tiers:

• Basic (free) — speech recognition + basic error correction + 1-row AI candidates
• Pro (from ¥12/mo) — 4-row AI candidates + terminology-aware translation + unlimited AI credits
• Ultra (from ¥15/mo) — everything in Pro + symbols/Markdown/snippets + 4-way candidates + Vibe Coding + SDD code-spec generation

The free tier gives 100 AI credits per day (1 AI call = 1 credit), resetting at midnight with no rollover. Subscribing to Pro or Ultra unlocks unlimited use. See the pricing page for details.

Free users get 100 AI credits per day (1 AI-enhanced recognition = 1 credit), reset automatically at midnight — no rollover to the next day. When you run out, AI-enhanced recognition pauses until the next midnight reset.

Pro or Ultra unlocks unlimited AI credits and all advanced features. There is no free trial — billing starts from the first cycle. Cancel anytime in System Settings; you keep access until the period ends.

The Windows app is free with no subscription: AI polishing and other enhancements use your own API key (BYOK), billed directly by the provider; offline recognition, meeting notes, to-dos, and exports need no key. See the Windows page.

Open System Settings → Apple ID → Subscriptions → find Kaikou → "Cancel Subscription". You keep the optional features until the period ends, then it automatically downgrades to Basic (free). Inserted text and your settings are unaffected.

It depends on your Family Sharing settings and the subscription type — the App Store subscription page is the source of truth. Check whether it's marked as Family-Sharing compatible there.

When upgrading via Apple IAP, Apple prorates your remaining time — you only pay the difference. The exact amount is computed by Apple and shown on the upgrade screen.

If you have a Kaikou promo code, redeem it inside the app:

• Open Kaikou → Menu-bar K icon → Settings → Account → Subscription
• Click "Redeem Promo Code"
• Enter the code you received and confirm

The discount applies immediately after redemption. Each code works once; the specific benefit (free duration, discount, etc.) follows the code's own rules.

Kaikou requires macOS 14.0 (Sonoma) or later, on both Apple Silicon (M1/M2/M3/M4) and Intel Macs. macOS 15.0+ is recommended for the best experience.

Kaikou pastes into the target app by simulating Cmd+V, then restores your original clipboard. If pasting fails, check whether that app accepts standard paste actions at all. For persistent issues, write to support@17kaikou.com.

Kaikou is a lightweight menu-bar app with a tiny idle footprint. Recording and recognition briefly use CPU and release it immediately after. The SenseVoice model takes ~894MB of disk; runtime memory depends on model inference. Apple Speech mode uses almost nothing extra.

Debug logs live at /tmp/kaikou_debug.log and rotate automatically (2MB cap). You normally never need them, but when contacting support, attaching the log helps us pinpoint issues faster.