Speech to Text, Instantly
AI enhancement, keyboard launcher — offline and private.
Lite — smaller build with Apple Speech + remote ASR, no large local models
Latest release
AI enhancement, keyboard launcher — offline and private.
Lite — smaller build with Apple Speech + remote ASR, no large local models
Latest releaseEverything you need for voice input on macOS
Zero telemetry, zero analytics, zero data collection. Pair with a self-hosted LLM (e.g. Ollama) and all your data — voice, text, vocabulary — stays entirely on your machine.
Five backends — Apple Speech (zero setup), FunASR (Chinese), MLX-Whisper (99 languages, Apple GPU), Sherpa-ONNX, and Whisper API. Real-time streaming overlay during recording.
LLM-powered proofreading, translation, and chain pipelines via any OpenAI-compatible API. Clipboard enhancement (Ctrl+Cmd+V) for selected text. Extended thinking visualization.
Personal vocabulary learns your domain terms from correction history. Conversation history enables topic continuity. The more you use it, the more accurate it gets.
Alfred-style search panel — apps, files, clipboard history, bookmarks, calculator with unit conversion, and a command palette. Snippet auto-expansion as you type.
Python plugin system with leader keys, hotkeys, event listeners, persistent storage, and shell commands for macOS automation.
Three simple steps — no window switching needed
Hold fn and speak. A live overlay shows partial results in real time. Cmd to restart, Space to cancel, Z for recent history.
Release the key — WenZi transcribes your speech and optionally enhances it with AI.
Result is typed into the active app, or shown in a preview panel to edit, switch modes (⌘1–⌘9), and compare before confirming.
The Preview panel — review, edit, switch modes, and compare models before confirming
Choose the best engine for your language and hardware
| Backend | Language | Speed | Accuracy | Streaming | Download |
|---|---|---|---|---|---|
| Apple Speech ⚠ | Multiple | Fast | Good | Supported | None (built-in) |
| FunASR Default | Chinese | Fast | High | No | ~945 MB |
| MLX-Whisper | 99 languages | Medium | High | No | 75 MB – 1.6 GB |
| Sherpa-ONNX | Multiple | Fast | High | Supported | Varies by model |
| Whisper API | Multiple | Depends on network | High | No | None (cloud) |
Download, drag, and go — ready in seconds
| Standard | Lite | |
|---|---|---|
| Local ASR | All 5 backends | Apple Speech (built-in) |
| Remote ASR | Whisper API | Whisper API |
| AI Enhancement | Full | Full |
| Launcher | Full | Full |
| Scripting | Full | Full |
| App Size | ~945 MB | ~64 MB |
| ⬇ Download | ⬇ Download |
Learn how to get the most out of WenZi