Free appOpen source

AI voice input.

Speak naturally. Turn your thoughts into clear text, right where you write.

macOS 14+ · Apple Silicon · Free app, provider usage billed separately

Download → Install → Add provider keys → Try your first dictation

Notes · Email · Chat

Speak → Transcribe → Polish → Insert

Enable JavaScript to play the interactive example. No microphone is used.

Your words. Straight to text.

Just want to dictate? Turn speech into text without asking AI to rewrite your wording.

Vowrite / DictationPreset example · no recording
“Let’s meet on Thursday at ten.”

…

SAY IT

A thought. A finished draft.

Speak naturally. Turn pauses and repeated words into a message that is ready to use.

VowriteInteractive example

Let’s meet Thursday at 10. Bring the updated design!

✨

Smart cleanup

Filler words gone, grammar fixed, punctuation added — and your personal vocabulary auto-corrects itself as you keep dictating.

⚡

Blazing fast

Speculative polish pre-warms connections during recording. Response time depends on the provider, network and recording length.

MAKE IT YOURS

The right words, wherever you write.

A quick reply, a thoughtful email, a tidy note. Try the same thought in three different places.

VowriteInteractive example

Hi team, Let’s meet Thursday at 10 to review the updated design. Please bring the latest version. Thank you!

🎯

Per-app Modes

Vowrite senses the app in front of you and switches Mode automatically — Slack stays casual, your editor stays code-flavored. Manual switches always win.

📑

Command templates

15 built-in commands — Improve, Fix Grammar, TL;DR, Make Professional, and more — applied straight to your selected text.

📄

Markdown & output styles

Built-in Markdown formatting plus 7 more Output Styles — bullet lists, headings, and code blocks land exactly right, per Mode.

04 / VOICE TRANSLATION

Speak one language. Write another.

Press the translation shortcut on Mac or choose Translate in the iOS keyboard. Vowrite recognizes speech and writes the requested output language into a compatible text field. Language support depends on your models.

Vowrite / TranslateInteractive example · no recording
🇨🇳 You say Chinese

嗯…那个,明天下午三点开个会,把翻译模式的设计方案过一下。

→
🇺🇸 Vowrite types English

Let's meet at 3 pm tomorrow to walk through the Translate mode design.

Two ways to trigger. Zero context switch.

Translate mode is a separate gesture, so it never overwrites the Mode you were already using.

macOS · Global hotkey

A second hotkey, just for translation

Press ⇧⌥ Space to start a translation recording. Stop with ⌥ Space or the overlay Stop control. Your current Mode (Email, Code, Note…) stays unchanged.

⇧ + ⌥ + Space
  • Configurable in Settings → Keyboard Shortcuts.
  • Conflict detection if it collides with the Dictate hotkey.
  • Recording overlay shows a target-language badge while you speak.
iOS · Keyboard arc

Long-press the mic, slide to translate

Inside the Vowrite keyboard, long-press the mic pill. Two arcs fan out — Dictate or Translate. Slide to one, release, speak. The keyboard shows a "Translating to {target}" banner in real time.

Dictate
Translate
  • Works in iOS apps and fields that allow third-party keyboards.
  • Selection haptic on tier change, success haptic on commit.
  • Configure source and target languages separately on each device. The iOS app shares these settings with its keyboard.
Language settings and the translation workflow

Set it once. Translate forever.

On Mac, open General → Language & Translation. On iOS, open Settings → Language & Feedback → Translation. Choose a source-language hint or automatic detection, then a target language. Each device remembers its own settings.

Need multiple language pairs? Create custom Translate modes — each with its own prompt, hotkey, and language pair. The quick settings only steer the built-in Translate mode, so custom presets stay untouched.

From voice to translation in one breath

A purpose-built pipeline that detects, recognises, translates, and types — without ever leaving your text field.

Speak naturally

Press the translation shortcut on Mac, or choose Translate from the iOS microphone menu. Speak in a language supported by your chosen models.

Auto-detect & transcribe

Compatible recognition models can detect the spoken language automatically. A configured source language provides a hint when the provider supports it.

LLM translates

A dedicated translation prompt rewrites the transcript into your target language — formal stays formal, casual stays casual.

Translation prompt

Type at the cursor

The finished translation is inserted into a compatible text field, such as a message, email or document.

Languages and everyday uses

Translation in both directions.

Choose a source and target language supported by your recognition and text models.

🇺🇸
English
en
🇨🇳
Chinese
zh-Hans
🇯🇵
Japanese
ja
🇰🇷
Korean
ko
🇪🇸
Spanish
es
🇫🇷
French
fr
🇩🇪
German
de
🇮🇹
Italian
it
🇵🇹
Portuguese
pt
🇷🇺
Russian
ru
🇸🇦
Arabic
ar
🇮🇳
Hindi
hi

Plus any other language your STT and LLM provider support — Whisper covers 90+, GPT and Claude translate even further.

Made for the moments your hands can't keep up

Real situations where pulling up Google Translate kills the flow.

Cross-language email

Your mother tongue is Chinese, your client only reads English. Speak the email in Chinese, and a polished English version appears in Mail.

"Hi James — wrapping up the v2 timeline now. Could you confirm the API contract by Friday? Happy to jump on a call this afternoon."

Multilingual team chat

Slack, WeChat, Telegram — speak naturally and reply in the channel's lingua franca. Your message lands clean, on time, on tone.

"Pushed the hotfix — translation prompt now ships fillers separately so STT noise doesn't leak into the output."

Travel & field trips

From a hotel in Tokyo, dictate in English — Vowrite types fluent Japanese into Notes, ready to flash to a stranger.

"すみません、新宿駅まではどう行けばいいですか?"

Reading & learning

Speak a sentence in your weaker language and check the English version. Or speak English to draft notes for a Spanish exam.

"Quiero practicar español con frases reales — Vowrite, traduce esto."
Compare translation tools

Why Vowrite Translate

Compare translated-text output, platform support and processing options to find the right workflow.

Feature Vowrite Apple Translate Google Translate DeepL Voice
Voice translation output Translate recordings into compatible text fields Spoken translation; copy/replace supported text Spoken translation; copy/share transcript Live voice/subtitles for meetings and conversations
App integration Mac text fields; iOS where third-party keyboards are allowed App and supported system text actions Web, iOS and Android Desktop/mobile/web; Teams, Zoom, Meet
Choose your own AI model Choose recognition and text models Apple-managed Google-managed Service-managed
API keys and subscriptions BYOK; provider costs separate Free Apple app Free consumer app Voice plans; eligible paid Translate plans include limited minutes
Custom translation modes & prompts Custom translation prompts Language/conversation controls Language/conversation controls Glossaries, terms and formality
Offline / local processing Local text model option; recognition needs a configured service Downloaded supported languages Downloadable languages; feature limits Online service; offline mode not documented
Open source Yes Not advertised Not advertised Not advertised

Checked 22 September 2026. Features vary by platform, model and plan. “Not documented” means no support claim was confirmed in the linked official documentation.
Apple Translate · Google Translate · DeepL Voice

05 / MAC + IPHONE

At your desk. In your pocket.

A Mac menu-bar app and an iPhone keyboard. The same voice workflow, with controls made for each device.

Ready to download

Vowrite for macOS

A menu-bar companion: press your shortcut, speak, then insert clean text into a compatible editable field.

macOS 14+ Apple Silicon

Insert text into the active editable field using clipboard paste. Accessibility permission is required; secure fields and apps that block paste may not work.

Want to test ahead? Try the v0.3.0.0-beta.1 pre-release →

macOS — installation & use

Install in 3 steps

1

Download the DMG

Download the latest macOS DMG from GitHub Releases. Sparkle verifies updates with EdDSA signatures; this does not imply Apple notarization.

Download Vowrite.dmg →

Trying the beta? v0.3.0.0-beta.1 is signed with a new code-signing identity and update key, so Vowrite 0.2.3.0 and earlier cannot update to it automatically. Download it once from the pre-release page, then grant Microphone and Accessibility permission again.

Download v0.3.0.0-beta.1 →
2

Drag to Applications

Mount the DMG, drop the app into /Applications. First launch will prompt for Microphone + Accessibility permission — grant both (the second enables Cmd+V keystroke output).

3

Add an API key

Open Settings → API Keys to add your recognition and text-provider keys, then select them under Models. Keys stay in Keychain at rest and are used to authenticate directly with providers. No Vowrite account is required.

Press ⇧⌥ Space to start translation. Stop with ⌥ Space or the overlay Stop button. Dictation supports hold-to-record when Push-to-Talk is enabled. Learn more →

What's in the Mac app

Global hotkey Carbon global shortcuts, including in fullscreen apps; avoid conflicting system shortcuts.
Cmd+V injection Clipboard paste into compatible editable fields in browsers, editors and other apps.
Floating overlay Choose from five recording indicators, including Orb Pulse. The beta adds 80 built-in Voice Bar effects that react to your microphone level.
Translate hotkey ⇧⌥ Space starts translation; use ⌥ Space or Stop to finish.
Custom Modes Email · Code · Slack — each with its own polish prompt
Auto-update EdDSA-signed Sparkle appcast, opt-in beta channel

Build from source

Vowrite for iOS

Speak from your iPhone keyboard. The companion app handles recording and refinement, then the keyboard inserts the result.

iOS 17+ A14+

The iOS app is open source and currently requires building in Xcode. No public App Store or TestFlight distribution is available.

iOS — source build & keyboard

VowriteIOS app

Hosts the engine, settings, history, Modes, and language pickers. Runs the recording service in the background so the keyboard can stay tiny.

⇄

VowriteKeyboard extension

A real iOS keyboard. Tap the mic to dictate, long-press to pick Translate. Talks to the host app over App Group + Darwin notifications.

Install (build from source)

1

Clone the repo

Requires Xcode 15+ and a free Apple developer account.

git clone https://github.com/Joevonlong/Vowrite.git
cd Vowrite/VowriteIOS
open VowriteIOS.xcodeproj
2

Sign and run

In Xcode → Signing & Capabilities, choose your own Team for both the VowriteIOS and VowriteKeyboard targets; the project does not preset one. If Xcode says a bundle identifier is unavailable, change com.vowrite.ios and com.vowrite.ios.keyboard to identifiers you own, and keep the keyboard ID a child of the app ID (for example com.example.vowrite and com.example.vowrite.keyboard). Also replace the App Group group.com.vowrite.shared with one you own, in VowriteIOS.entitlements, VowriteKeyboard.entitlements and appGroupID in VowriteStorage.swift. Then plug in your iPhone, hit ▶ and deploy both targets.

3

Enable the keyboard

iOS Settings → General → Keyboard → Keyboards → Add New Keyboard → Vowrite. Tap "Vowrite" again and toggle Allow Full Access (required for clipboard + network).

Long-press the mic and you'll see the Dictate / Translate arc — slide up to choose, release. See the gesture →

What's in the iOS app

Real keyboard Works in iOS apps and fields that allow third-party keyboards.
Long-press arc Hold the mic, slide up to pick Dictate or Translate. Release to fire.
Translate banner "Translating to English" sticker stays on screen so you never wonder what's about to come out
Mode editor parity Edit Modes and source/target languages in the iOS app.
History and text deletion Review transcripts and swipe to delete history. The keyboard can delete a word, line, paragraph or all available text.
Keychain + App Group Keys are stored in Keychain and used to authenticate directly with providers. The app and keyboard communicate through App Group and Darwin notifications.

Set up each device separately. Mac and iPhone settings do not sync automatically.

Shared engine and platform architecture

One engine, two platforms

Both apps share VowriteKit: audio, provider integrations, Modes and translation. On iOS, the app and keyboard share settings through an App Group on the same device. Mac and iOS settings are separate; there is no automatic cross-device sync.

🧠
VowriteKit Shared Swift library — audio, Whisper, polish, Modes, translation
↗
🖥️
VowriteMac Menu bar UI, Carbon hotkey, CGEvent injection, Sparkle
↘
📱
VowriteIOS + VowriteKeyboard Host app + keyboard extension, App Group IPC, haptics

A little setup. All yours.

Choose what listens and what helps you write. Your preferences stay in your hands.

Your providers. Your choice.

Cloud + Local

Choose providers independently for recognition and refinement. Combine cloud services with local models.

Choose both providers to see your combination.

No providers added yet.

    Illustrative setup only. Never enter a real API key here.

    Local text refinement is available through Ollama or MLX Server on Mac. Built-in Sherpa speech recognition is currently unavailable; a cloud recognition provider is needed for this example.

    Shortcuts, languages and writing styles
    VowriteInteractive example

    OpenAI · Groq · Google · Anthropic · DeepSeek · Ollama

    • OpenAI
    • Groq
    • DeepSeek
    • Gemini
    • Claude
    • Deepgram
    • iFlytek
    • MLX Server
    • Ollama
    • Cerebras
    • OpenRouter
    • Together AI
    • Volcengine
    • Qwen
    • SiliconFlow
    • Kimi
    • MiniMax
    • Zhipu
    Full provider guide

    Your AI, your rules

    Pair any speech-to-text engine with any polish model. Bring your own keys.

    ★ Recommended combo
    Speech → Text
    Groq
    whisper-large-v3-turbo
    Cloud recognition · provider limits apply
    Polish & Format
    DeepSeek
    deepseek-flash
    Text refinement · usage billing

    Example combination: Groq for speech recognition and DeepSeek for refinement. Check each provider’s current prices, account access and limits; latency depends on your recording and network.

    Or pick from 20+ providers

    OpenAI
    gpt-4o-mini-transcribe · gpt-5.4-mini
    STTPolish
    All-in-One
    Deepgram
    Nova-3 · WebSocket
    STT
    36+ Languages
    Gemini
    gemini-3.8-flash
    Polish
    Google
    Claude
    claude-sonnet-5 · SSE
    Polish
    Anthropic
    iFlytek
    23 Chinese dialects
    STT
    Dialect Support
    MLX Server
    Apple Silicon · local text refinement
    Polish
    Local text model
    Ollama
    Local text refinement
    Polish
    Local refinement
    Cerebras
    gpt-oss-120b
    Polish
    Provider plan
    More provider integrations
    OpenRouter · Together · Volcengine · Qwen · SiliconFlow · Kimi · MiniMax · Zhipu
    STTPolish
    Your Server

    Free app. Pay for what you use.

    Vowrite is free and open source. Cloud usage is billed directly by the providers you choose — not a Vowrite subscription.

    How free & usage works →

    Why Vowrite

    See how it compares.

    Feature Vowrite Typeless Superwhisper macOS Dictation
    Price Free app; provider usage billed separately Free 8,000 words/week; Pro $12/mo yearly or $30 monthly Free tier; paid Pro plans Included with Apple devices
    Multiple AI providers Choose recognition and refinement providers Service-managed models Multiple cloud/local models Apple-managed
    Bring your own API key Yes Not documented BYOK with Pro Not applicable
    Data processing and history Local history; chosen providers process audio/text Cloud processing; local history, optional sync Local or cloud, model-dependent On-device or server; check settings
    Local processing Local refinement; bundled offline recognition unavailable Cloud transcription On-device speech models Device/language-dependent
    Mac + iOS Mac download; iOS source build Mac + iOS Mac + iOS, also Windows/Android Supported Apple devices
    Open source Yes Not advertised Not advertised Proprietary system feature
    AI text cleanup Optional chosen text model Yes Model/mode-dependent Punctuation; rewriting is separate
    Chinese + English mixed Provider/model-dependent Mixed languages supported Model-dependent Select/switch dictation languages
    Text replacement & auto-learning Replacement rules; correction learning on Mac Dictionary + personalization Vocabulary + replacements System replacements in supported apps
    Dictation history Local searchable history Local, optional cloud sync Local history No transcript library documented
    Custom hotkey Yes (Mac) Yes (desktop) Yes (desktop) Yes (Mac)

    * Typeless free tier: 8,000 words/week.

    Checked 22 September 2026. Features vary by platform, model and plan. “Not documented” means no support claim was confirmed in the linked official documentation.
    Typeless · Superwhisper · Apple Dictation

    Why we are building Vowrite

    Our Vision

    We believe input should be invisible. You shouldn't think about how to input — you should just think, speak, and your ideas arrive exactly as you mean them. Vowrite is the bridge between your mind and the screen.

    Our Ambition

    Today, Vowrite is a smart voice keyboard. Tomorrow, it's your input layer — adapting your words to any context, connecting to any workflow, speaking your language in every sense. Voice is just the beginning.

    Our Goal

    Make input effortless for everyone. No learning curve. No configuration headache. Just speak, and your ideas arrive — clean, precise, and unmistakably yours.

    A few things to know

    How does Vowrite turn my voice into clean text?

    Vowrite transcribes your recording, then optionally uses a text model to remove fillers, fix grammar and add punctuation. It pastes the result at the cursor. Processing time depends on the models, network and audio length.

    Is Vowrite free to use?

    Yes. Vowrite is free and open source. Bring your own provider credentials; cloud usage is billed directly by your chosen providers under their current rates and limits.

    Does it work in any app on Mac?

    Insert text into the active editable field using clipboard paste. Accessibility permission is required; secure fields and apps that block paste may not work.

    Can I mix Chinese and English in one sentence?

    Mixed Chinese and English is supported by compatible recognition models. Quality and language coverage depend on the selected model; no manual language switch is needed when automatic detection is supported.

    What AI providers are supported?

    The registry includes 20+ provider integrations. OpenAI, Groq, Deepgram and iFlytek offer recognition; DeepSeek, Gemini and Claude offer text refinement. Ollama and MLX Server provide local text refinement. Built-in Sherpa recognition is currently unavailable.

    Is my data private?

    History is stored locally. Cloud recognition sends audio to your chosen recognition provider; cloud refinement sends the transcript and any enabled prompt context to the text provider. Their retention policies apply. Local text models keep that refinement step on your machine.

    How do I get the newest models?

    App updates include reviewed provider presets. Where supported, refresh the model list in provider settings, or enter a compatible model ID. A provider catalog listing is not a guarantee of account access or compatibility.

    Coming soon

    Voice-triggered search — Ask questions by voice and get instant answers without leaving your app.

    iOS App Store distribution — The iOS keyboard is open source and buildable today — official App Store and TestFlight distribution is planned.

    Ready to stop typing?

    🔑 API Key (OpenAI or other) 🎤 Microphone permission ♿ Accessibility for automatic paste

    Or build from source: cd Vowrite/VowriteMac && swift build
    View source on GitHub

    v0.2.3.0 · macOS 14 Sonoma · Apple Silicon

    Want to test ahead? Try the v0.3.0.0-beta.1 pre-release →

    Build from source: read the GitHub guide