Input 0 — Zero typing. Pure voice.

Input 0 is a macOS voice-to-text transcription tool built with Rust and Metal that captures speech via a global hotkey, transcribes it locally using GPU-accelerated AI models, polishes the text with an LLM, and auto-pastes the result into any active application.

What is Input 0?

Input 0 is a macOS application (Apple Silicon only) that converts spoken input into polished written text. It takes voice audio from your microphone, transcribes it locally using one of six AI engines (including Whisper, SenseVoice, and Paraformer), optionally passes the text through a user‑owned LLM API for grammar and filler‑word removal, then automatically pastes the result into the currently focused window. The app is developed by the team behind the GitHub repository 10xChengTu/input0.

Key Features

• Zero‑typing workflow — Press and hold ⌥ Space from anywhere in macOS, speak, then release to transcribe and paste text instantly. No window switching.
• Local transcription via Metal GPU — All speech‑to‑text processing runs on your Mac’s GPU using Metal acceleration; audio never leaves the device.
• AI text polishing — An LLM (connected via your own API) removes filler words like “um” and “uh”, fixes grammar, and corrects 40+ technical terms pronounced by sound (e.g. “rui en te” → “React”). Uses a 10‑message history for context.
• Auto‑paste integration — Transcribed and polished text is injected directly into Slack, VS Code, browsers, or any other text field without extra clicks.
• Multi‑language support — 6 AI engines with 12 models covering 99+ languages. Smart model recommendation picks the best option for the detected language. Supports English, Chinese, Japanese, Korean, Spanish, French, German, and more.
• Custom vocabulary — Add your own terms (names, brands, jargon) that the AI learns to transcribe correctly. Corrections are validated by the LLM.

Who is it for?

• Developers and engineers — Dictate code comments, commit messages, or documentation directly into VS Code or a terminal without switching context.
• Writers and content creators — Capture rough ideas by voice, have them automatically cleaned of filler words, and paste into a text editor or note app.
• Multilingual professionals — Transcribe in any of 99+ languages using dedicated CJK‑optimized engines (SenseVoice, Paraformer, FireRedASR) without changing tools.

What can you do with Input 0?

• Dictate meeting notes: Speak “Let’s schedule a meeting for next Tuesday to discuss the new React system benchmarks” while in a calendar app; the polished text appears directly in the event description.
• Transcribe and translate: Use the English Moonshine model (5× faster) for quick transcriptions, or switch to a Japanese‑optimized model for bilingual workflows.
• Add custom jargon: Pre‑load technical terms like “Kubernetes” or “CI/CD pipeline” so the AI transcribes them correctly without manual correction.

How does Input 0 work?

The workflow has three steps: (1) Hold ⌥ Space from any app – a minimal floating pill appears near the cursor. (2) Speak your thoughts – release the key to trigger local transcription via a Metal‑accelerated AI engine. (3) The transcribed text is polished by a user‑provided LLM, then automatically pasted into the active application.

Pricing

Input 0 is offered under a freemium model. Basic usage (local transcription with the default Whisper model, limited custom vocabulary) is free. Additional AI engines, higher‑accuracy models, and LLM polish require a paid upgrade (exact tier names and prices are not listed on the page).

FAQ

Is Input 0 free?

Basic local transcription with Whisper and a limited custom vocabulary is free. Access to additional AI engines (SenseVoice, Paraformer, Moonshine, etc.) and LLM‑based text polishing requires a paid subscription.