Open Source · Privacy First · 100% Local

TypeIn

Open-source AI voice input.
Every byte stays on your Mac.

Press a key, speak, release. Your words appear at the cursor — instantly, accurately, privately. No cloud, no subscription, no data ever leaves your device.

Download for macOS View on GitHub

How It Works

Three steps. Zero friction.

No setup wizards, no account creation. Install, press a key, and start talking.

Hold

Press fn to start recording. A floating indicator appears so you always know it's listening.

Speak

Talk naturally in any of 7 languages. On-device AI transcribes your speech in real time.

Done

Release the key. Text appears exactly where your cursor is — in any app, any text field.

Hold 5+ seconds for long dictation mode. Works in every app — editors, browsers, chat, terminal.

Features

More than just dictation.

TypeIn handles everything from a quick message to a two-hour meeting.

🎙

Meeting Mode

Record an entire meeting with one click. TypeIn automatically detects when someone is speaking, transcribes each segment with timestamps, and saves the full audio. When the meeting ends, polish the transcript with one-click AI cleanup.

✨

AI Text Cleanup

Raw speech recognition is good. With optional AI post-processing, it's great — filler words removed, punctuation fixed, homophones corrected. Use a local AI or cloud models. Off by default, only text is sent, never audio.

📋

Full History

Every recording is searchable, replayable, and shows both raw and cleaned text side by side. Double-click to make your own correction. Export everything as a zip. Auto-delete after 7, 30, or 90 days — your choice.

🌐

7 Languages

Full interface and transcription support for English, Chinese, Japanese, Korean, Spanish, French, and German. Language auto-detection or manual selection. Especially optimized for Chinese accuracy.

Privacy First

Your voice never leaves your Mac.

Not "we don't sell your data." We never even see it. Everything runs on your device, and the code is open for anyone to verify.

🔒

100% On-Device

Speech recognition runs entirely on your Mac. No server, no upload, no exception.

📦

Open Source

Every line of code is on GitHub. No hidden analytics, no tracking. Audit it yourself.

🗄

Local Storage

Recordings, history, and settings stay in your Mac's local folders. API keys secured in macOS Keychain. Export or delete anytime.

⚡

Cloud is Optional

AI text cleanup can use cloud models, but it's off by default and only sends text — never audio. Core voice recognition works offline.

For Developers

Local ASR, your way.

TypeIn includes a built-in API server and CLI tool. Use speech-to-text from your code, scripts, or AI agents — no cloud API key needed.

📡 API Server

                    # OpenAI-compatible, drop-in replacement

                    curl http://localhost:48080/v1/audio/transcriptions \

                      -F file=@recording.m4a

                    # Works with any OpenAI SDK

                    from openai import OpenAI

                    client = OpenAI(base_url="http://localhost:48080/v1")

                    result = client.audio.transcriptions.create(

                      model="qwen3-asr",

                      file=open("meeting.m4a", "rb")

                    )

💻 CLI Tool

                    # Transcribe any file from your terminal

                    typein meeting.mp4

                    # Generate subtitles

                    typein lecture.mp4 --format srt > subtitles.srt

                    # Check server status

                    typein status

Endpoints

POST /v1/audio/transcriptions
GET /v1/models
GET /health

Formats

All major audio and video formats. Non-native formats auto-converted via built-in codec. Per-request model and language selection.

Security

Optional Bearer token auth. localhost-only by default. LAN access opt-in.

Open-source AI voice input.Every byte stays on your Mac.