Skip to content

Speech to text,
BYOK

Press a shortcut, speak naturally, get text in any app. Go fully local for privacy or bring your own API keys (BYOK) for cloud accuracy.

Free & open source · macOS 12+

Everything you need to dictate

Local, cloud, or both. No subscriptions.

Local or Cloud — You Choose

Run transcription fully on-device for privacy, or bring your own API keys for cloud-powered accuracy. Flexible by design.

Press & Speak

One keyboard shortcut to start. Hold, toggle, or always-on — speak naturally, get text instantly in any app.

LLM Polish

Clean up transcriptions with OpenAI, Anthropic, Gemini, Groq, OpenRouter, Cerebras, Z.AI, Apple Intelligence, or your own endpoint. Mild, medium, or aggressive — your call.

Dictionary & History

Add custom terms so your jargon lands right. Full transcription history with daily stats and words-per-minute tracking.

Native & Fast

Built with Tauri and Rust. Lightweight, instant startup, native macOS feel. Voice activity detection stops recording when you stop speaking.

How it works

Pick your engine, choose a post-processing model, set your polish level — done.

Your voice

Transcription

Post-process

Polish mode

Polished text

See it in action

Watch the app in action. Try the Fn key below to see real-time captioning.

General
Shortcuts
Models
Dictionary
Polish
History

Transcription Models

Select a transcription model or download additional models.

My ModelsLibrary

OpenAI

OpenAI’s cloud speech-to-text API. Fast and accurate with support for 57+ languages.

Verify

Uses a WebSocket connection instead of file upload for lower latency

Language
Auto detect
Leave empty to auto detect
Glossary:
Temperature

Higher values produce more random results (0-1). Only supported by whisper-1.

0
Multi-languageTranslate

Deepgram

Deepgram’s Nova speech-to-text API. Fast and accurate with 50+ language support.

Multi-language

Parakeet V2

Active

English only. The best model for English speakers.

English Only

API Configuration

Anthropic · claude-sonnet-4-20250514

Prompts

Mild - Correct Transcript
Built-in
Medium - Improve Fluency
Built-in
Aggressive - Restructure & Format
Built-in
You are a transcript-to-prose converter. Process the transcript into polished written prose. Fix spelling, capitalization, and punctuation. Remove filler words and stutters. Group sentences into paragraphs by rhetorical function. When the speaker lists parallel items, format them as a numbered list.
Parakeet V2
v0.1.13

Press & Speak

Hold the Fn key, speak naturally, get text instantly

Choose your engine

5 local engines. Pick what works for your language and speed.

EngineVariantsLanguagesNote
WhisperSmall, Medium, Turbo, Large99+ languagesMost popular
ParakeetV2, V3English & 25 EuropeanNVIDIA
MoonshineBase, V2 Tiny, V2 Small, V2 MediumEnglishUltra-fast
SenseVoiceInt8CJK, Cantonese, EnglishMultilingual
Breeze ASRStandardTaiwanese MandarinSpecialized

Prefer cloud? Bring your own OpenAI or Soniox API key for cloud-powered transcription with real-time streaming.

Start dictating today

No account needed. Download and start speaking.