What should you check first in this article?

Moonshine Voice is a local speech-recognition toolkit that runs on a CPU. Its English CLI updates text while listening to a microphone; the current Korean Tiny model is non-streaming.

When is it better not to choose this?

The Moonshine model list distinguishes MIT-licensed English streaming models from the Korean Tiny model under the Community License. That Community License is non-commercial, so do not assume the checkpoint can be used in a commercial service or paid product. First decide whether you need live English or offline Korean recognition, then choose the matching model and license.

Where can you check the conditions and sources?

The Moonshine model list distinguishes MIT-licensed English streaming models from the Korean Tiny model under the Community License. That Community License is non-commercial, so do not assume the checkpoint can be used in a commercial service or paid product. First decide whether you need live English or offline Korean recognition, then choose the matching model and license.

Local speech recognition & synthesis

Transcribe English speech in real time with Moonshine on CPU

English can use a live streaming model, while the current Korean model has a different streaming path and license.

Moonshine Voice is a local speech-recognition toolkit that runs on a CPU¹. Its English CLI² updates text while listening to a microphone; the current Korean Tiny model is non-streaming.

Start with live English input

Moonshine Voice is a local speech recognition toolkit that turns speech into text. It runs on a CPU, and the English microphone CLI updates the transcript while you speak. If you want an on-device workflow that does not send audio to a server for transcription, begin with a short microphone test. Moonshine Voice documentation

Run in Windows PowerShell

python -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install moonshine-voice
moonshine-voice mic --language en

The official quickstart installs moonshine-voice, then opens the English microphone with mic. Activate a virtual environment³ for your operating system and install the package. The first run prepares model assets, so it needs an internet connection and local storage.

Prepare a virtual environment and Moonshine CLI

python -m venv .venv
source .venv/bin/activate
python -m pip install moonshine-voice
moonshine-voice mic --language en

An audio waveform displayed on a laptop beside a microphone — Allow microphone access, then speak a short phrase in English.

Check microphone access and live text

After the operating system grants microphone access to the terminal, spoken words should appear. Say a short phrase and pause to see whether the transcript adds text or finishes a line. If nothing appears, enable microphone permission for the terminal in system privacy settings and check that the CLI language is en. Official quickstart

To add transcription to a web app, use @moonshine-ai/moonshine-wasm. Start with the documented package or CDN example, then connect model downloading and microphone permission.

The Korean model follows a different path from live English

Current model documentation does not list Korean among streaming languages; it lists a separate 26M-parameter Tiny non-streaming model. The English CLI’s live workflow therefore does not carry over to Korean dictation. If you need Korean recognition, test it with your actual speech first. Moonshine model list

The Moonshine model list distinguishes MIT-licensed English streaming models from the Korean Tiny model under the Community License. That Community License is non-commercial, so do not assume the checkpoint⁴ can be used in a commercial service or paid product. First decide whether you need live English or offline Korean recognition, then choose the matching model and license.

A desk with headphones, a microphone, and a transcription workspace — Review the transcript as it changes while you speak.

Decide whether live updates fit your task

Read a short passage from an English meeting and compare names and numbers on screen with the original audio. Paste accurate lines into a notes draft, and replay important names before using them in the minutes.

A recorder, notes, and headphones for reviewing speech recognition — Check the model list to see whether your language has live support.

Terminology notes

CPU — The central processor that runs the operating system and general-purpose instructions; in AI workloads it also handles tasks such as preprocessing and data movement.
Back to the text
CLI — Short for Command-Line Interface: operating a program by entering commands in a terminal.
Back to the text
Python virtual environment — An isolated space for installing Python packages per project, helping reduce package-version conflicts.
Back to the text
Checkpoint — A file containing saved model weights and related state. Versions or tasks in one model family may use different checkpoints.
Back to the text

Frequently asked questions

What should you check first in this article?: Moonshine Voice is a local speech-recognition toolkit that runs on a CPU. Its English CLI updates text while listening to a microphone; the current Korean Tiny model is non-streaming.
When is it better not to choose this?: The Moonshine model list distinguishes MIT-licensed English streaming models from the Korean Tiny model under the Community License. That Community License is non-commercial, so do not assume the checkpoint can be used in a commercial service or paid product. First decide whether you need live English or offline Korean recognition, then choose the matching model and license.
Where can you check the conditions and sources?: The Moonshine model list distinguishes MIT-licensed English streaming models from the Korean Tiny model under the Community License. That Community License is non-commercial, so do not assume the checkpoint can be used in a commercial service or paid product. First decide whether you need live English or offline Korean recognition, then choose the matching model and license.