DuckTypeDuckType

Everything DuckType can do

100+ languages, CJK romanization, regional spelling corrections, AI skill chaining, and a lot more.

Transcription

9 cloud providers, 3 local model families

Pick the provider that fits your needs. Use your own API key for full control, or subscribe to DuckType Cloud and we handle it for you.

Cloud providers

DuckType Cloud

Managed

OpenAI

Whisper

Groq

Whisper

Deepgram

Nova

Mistral

Fireworks

Whisper

Cloudflare

Workers AI

Baseten

ElevenLabs

Scribe

Local models (offline)

Whisper

100 languages, multiple sizes

Parakeet TDT

English, high accuracy

SenseVoice

Chinese, Japanese, Korean, English, Cantonese

100+ languages

From English and Spanish to Japanese, Arabic, and Hindi. Language support varies by provider, with up to 100 languages on some engines.

Bring your own key

Use your existing OpenAI, Deepgram, or Groq account. No markup on API costs. Your keys are stored locally, never sent to our servers.

Or let us handle it

DuckType Cloud gives you 200,000 words/month with automatic provider fallback. Works on any device, no API keys to manage. Just start dictating.

Recording Modes

Six ways to start dictating

From push-to-talk for quick edits to always-on auto mode for continuous dictation. Every mode works globally, even when DuckType is minimized.

Push-to-talk

Hold a key to record, release to transcribe. The classic mode for precise control.

Click to record

Toggle recording with a click or keyboard shortcut. Good for longer dictation sessions.

Double-tap

Double-tap a modifier key to start recording. Quick activation without reaching for a shortcut.

Fn key hold

Hold the Fn key to record. Native feel, no custom shortcut needed.

Auto mode

Always-on listening with voice activity detection. Sentences are segmented by silence gaps. The microphone automatically switches to your preferred device when it becomes available.

Instant recording

Reuses the microphone stream between recordings so there is near-zero activation latency. No Bluetooth warm-up delay. Recording starts the instant you speak.

AI Processing

Transform text after transcription

AI skills run on your transcription to fix grammar, translate, summarize, or do anything you can describe in a prompt.

Custom AI skills

Create skills with custom prompts. Fix grammar, translate to another language, summarize meeting notes, rewrite for tone, or anything else. Skills can run automatically on every transcription or be triggered manually.

Skill chaining

Chain multiple skills in sequence. The output of one becomes the input of the next. Transcribe, then translate, then format as bullet points, all in one pass. Save chains as presets for different workflows.

7+ LLM providers

Skills work with your choice of language model. Use cloud APIs or run locally with Ollama for fully offline AI processing.

OpenAI (GPT)Anthropic (Claude)Google (Gemini)GroqOpenRouterOllama (local)Any OpenAI-compatible

Import audio & video

Drag and drop, paste, or pick any audio or video file. MP4, MOV, MP3, WAV, FLAC, OGG, WebM, and 25+ more formats. DuckType extracts the audio and converts it to text. Skills run on the result just like live dictation.

Language Intelligence

Beyond transcription accuracy

DuckType understands regional spelling variants, romanizes CJK scripts, and learns your vocabulary. No other dictation app does this.

CJK Romanization

Dictate in Japanese, Chinese, or Korean and get romanized Latin-script output alongside the original text. Useful for language learners, subtitlers, and anyone working across writing systems.

東京

tōkyō

JapaneseRomaji

Lindera tokenizer for accurate kanji readings

你好世界

nǐ hǎo shì jiè

ChinesePinyin

Character-level pinyin with tone marks

한국

han gug

KoreanRevised Romanization

Hangul decomposition using standard system

Regional spelling corrections

Most transcription engines output American English or Brazilian Portuguese by default. DuckType automatically corrects spelling to match your regional variant.

British English

colorcolour
analyzeanalyse
centercentre

European Portuguese

bebêbebé
abdômenabdómen

Swiss German

straßestrasse

Smart dictionary

Custom word lists

Create multiple dictionary lists for different contexts. Technical terms, product names, medical vocabulary. Toggle lists on and off as needed.

Text replacements

Define shortcuts that expand into longer text. Type abbreviations, email signatures, code snippets, or frequently used phrases.

1,900+ emoji

Built-in Unicode emoji library. Define emoji shortcuts for quick insertion during dictation.

Productivity

Built for people who dictate all day

Global shortcuts, deep customization, and a progression system that keeps you motivated.

Global shortcuts

Every action has a configurable global hotkey. Push-to-talk, toggle skills, open scratchpad, paste last transcription. Remap anything to fit your workflow.

Paste at cursor

Transcribed text is pasted directly where your cursor is. Works in any app: text editors, browsers, chat windows, terminals.

Statistics & levels

Track words per minute, daily word counts, and dictation streaks. Hit milestones and level up your duck from Duckling to Admiral.

Skill presets

Group multiple skills into reusable presets. Switch between workflows with a single shortcut. One for emails, one for code comments, one for meeting notes.

CLI

Command-line interface for power users. Transcribe files, manage settings, and integrate DuckType into scripts and automation workflows.

Deep customization

Per-provider model selection, custom AI prompts, personal dictionary with shortcuts and templates. Tune every detail to match how you work.

Privacy

Your data stays yours

DuckType is designed so your data never goes anywhere you didn't choose. No surveillance, no telemetry by default, no data harvesting.

No screen reading

DuckType never reads your accessibility tree, captures window contents, or inspects what's on your screen. Accessibility access is optional and only used for cursor positioning.

No URL logging

DuckType does not track which apps you use, which websites you visit, or what you're doing when you dictate. Zero behavioral data is collected.

Fully offline option

Use local models like Whisper, Parakeet, or SenseVoice for transcription and Ollama for AI skills. Nothing leaves your device. No internet connection required.

Independent, not VC-backed

DuckType is independently built. No investors pushing for growth metrics or data collection. Your subscription pays for development. That's it.

Technical

Built with Rust, not Electron

DuckType uses Tauri and Rust for native performance with a fraction of the resource usage of Electron-based alternatives.

Tauri + Rust

Native Rust backend with a lightweight webview frontend. No bundled Chromium. Low memory and CPU footprint.

~30 MB download

The full app is under 30 MB. Compare that to 238 MB for Electron-based competitors.

SQLite storage

Full transcription history stored locally in SQLite. Fast search, edit, and re-transcribe from history at any time.

macOS 11+

Supports macOS 11 Big Sur and newer. Android, Windows, and Linux support is coming soon.

Try DuckType

Download for macOS 11 Big Sur and newer. Bring your own API key or use offline models. No account or credit card needed.

Download for macOS