Speak, and it becomes a prompt

What Whispree does

Press a shortcut → speak → stop, and Whispree handles dictation (STT) → (optional) AI correction → automatic paste into the app you were last using in one go. Both the dictation engine and the correction AI are swappable at runtime, and everything works end to end with local models alone.

Stay in your flow

Default shortcut Ctrl+Shift+R. Press, speak, and the result lands in the previously focused app (editor, terminal, browser).

Local-first, your choice

STT runs on WhisperKit (local) · Groq (cloud) · MLX Audio (local); correction on local MLX · OpenAI · Groq · Claude subscription.

Dictation that never stops

Start the next recording immediately while the previous one is still processing. Insertion follows the order you spoke (FIFO).

Understands context

Capture the screen during recording and pass it to a vision model, or remember the browser tab and terminal session and paste back into exactly that spot.

Core features at a glance

Dictation & multi-recording Recording modes (push-to-talk / toggle), multi-recording queue, shortcuts, URL scheme.

STT engines WhisperKit · Groq · MLX Audio, language selection, silence auto-skip (VAD).

AI correction None · local MLX · OpenAI · Groq · Claude subscription, 4 correction modes.

Dictionary & Quick Fix Domain word sets, on-the-fly correction, iCloud sync.

Visual context & restore Screenshots for vision, browser tab and iTerm2/tmux restore.

Models & compatibility Local model registry, 'Can I Run' compatibility grades, downloads.

Requirements

• macOS 14 or later, Apple Silicon (arm64) only — Intel Macs are not supported.
• Microphone permission is required. Automatic paste needs Accessibility permission.
• Local models download on first use (WhisperKit STT ~1.5 GB, etc.). Cloud engines (Groq/OpenAI) work immediately once authenticated.

For detailed installation, permissions, and your first dictation, see Get started.