How popular is moonshine on GitHub?

moonshine-ai/moonshine has 11,173 stars and 619 forks on GitHub, and gained 38 stars in the last 7 days.

What license does moonshine use?

moonshine-ai/moonshine is released under the Other license.

moonshine-ai/moonshinePublic

Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces

AI summary: A fast, cross-platform family of open speech-to-text models designed specifically for low-latency live voice interfaces.

Star on GitHub

Stars: 11.2K; +14 today
Forks: 619
Watchers: 66
Open issues: 8
Open PRs: 2
Contributors: ~20
Commits: 666
Branches: 14

C++OtherCreated Oct 4, 2024Last push 2d agoLatest release v0.1.5+38 stars this week+177 this month

Quick answers

What is moonshine?: A fast, cross-platform family of open speech-to-text models designed specifically for low-latency live voice interfaces.
What does moonshine do?: Moonshine offers a suite of highly optimized speech-to-text (STT) and text-to-speech (TTS) models tailored for real-time applications. Unlike Whisper, which relies on fixed 30-second processing windows and recomputes from scratch, Moonshine utilizes an efficient streaming architecture that dynamically processes audio with caching to eliminate redundant computations. This approach achieves latencies often 5x lower than Whisper, making it suitable for live interactions and resource-constrained edge devices. It provides pre-built libraries and native examples across a wide array of platforms including iOS, Android, Linux, macOS, Windows, and Raspberry Pi, while offering multi-language support.
Who is moonshine for?: Moonshine is built for developers, roboticists, and mobile engineers who need to integrate responsive, offline speech recognition and synthesis into applications where Whisper's latency or edge platform fragmentation is a bottleneck.
How do I get started with moonshine?: pip install moonshine-voice moonshine-voice mic --language en
How popular is moonshine on GitHub?: moonshine-ai/moonshine has 11,173 stars and 619 forks on GitHub, and gained 38 stars in the last 7 days.
What license does moonshine use?: moonshine-ai/moonshine is released under the Other license.

Star history

since Jul 29, 2026

11.2K stars as of Oct 2, 2026. Measured daily since Jul 29, 2026; GitHub no longer exposes earlier star timestamps.

Contribution activity

commits per day, last 52 weeks

660 commits in the last yearLessMore

Signals and awards

derived from tracked data

Widely adopted
11,173 stars
Very active
660 commits in 52 weeks
Continuous integration
Automated checks passing
Repeat trending
4 trending appearances

What moonshine does

Moonshine offers a suite of highly optimized speech-to-text (STT) and text-to-speech (TTS) models tailored for real-time applications. Unlike Whisper, which relies on fixed 30-second processing windows and recomputes from scratch, Moonshine utilizes an efficient streaming architecture that dynamically processes audio with caching to eliminate redundant computations. This approach achieves latencies often 5x lower than Whisper, making it suitable for live interactions and resource-constrained edge devices. It provides pre-built libraries and native examples across a wide array of platforms including iOS, Android, Linux, macOS, Windows, and Raspberry Pi, while offering multi-language support.

Moonshine is built for developers, roboticists, and mobile engineers who need to integrate responsive, offline speech recognition and synthesis into applications where Whisper's latency or edge platform fragmentation is a bottleneck.

Dynamic input windowing: Processes audio without the rigid 30-second constraints of Whisper, significantly cutting down wasted computation on short phrases.
State caching: Reuses previous computations during streaming audio input to ensure near-instant feedback for live voice interfaces.
Cross-platform edge support: Ships with self-contained, prebuilt libraries and starter projects for iOS, Android, Raspberry Pi, macOS, Windows, and Linux.
Multi-language availability: Supports languages including English, Spanish, Mandarin, Japanese, Korean, Vietnamese, Ukrainian, and Arabic natively.
Variable model sizes: Offers Tiny, Small, and Medium streaming variants to allow developers to trade off parameter count (34M to 245M) for speed versus accuracy.

Where teams use it

Live voice assistants

Developers embed the streaming models to power interactive dialogue agents that require sub-200ms latency to feel natural.

On-device transcription

Mobile engineers build offline dictation apps for iOS and Android using the provided lightweight C++ native libraries.

IoT and embedded systems

Makers and roboticists integrate Moonshine onto Raspberry Pi devices using the optimized Python package to handle voice commands offline.

Desktop dictation tools

Windows and macOS application developers implement continuous speech recognition without incurring cloud API costs or data privacy concerns.

Getting started: pip install moonshine-voice moonshine-voice mic --language en

README

main branch

Moonshine Voice

Voice Interfaces for Everyone

Moonshine Voice is an open source AI toolkit for developers building real-time voice agents and applications.

Full guides, models, and API reference are at moonshine-voice.readthedocs.io.

Everything runs on-device — fast, private, and with no account or API keys.
Optimized for live streaming, with low latency by doing work while the user is still talking.
Speech to text models trained from scratch, from higher accuracy than Whisper Large V3 down to tiny 1MB models.
One library across Python, JavaScript/WASM, iOS, Android, macOS, Linux, Windows, and Raspberry Pi.

Quickstart

pip install moonshine-voice
moonshine-voice mic --language en

Every other platform is covered in the Quickstart, with runnable samples in Examples.

Building with a coding agent? Copy .agents/skills/moonshine-voice/ into your project's .agents/skills/ folder, or run npx skills add moonshine-ai/moonshine --skill moonshine-voice. The skill teaches the current API shape so the agent does not reach for Whisper or the old DialogFlow names.

Documentation

Quickstart to install and run on your platform.
Using the Library for transcription, text to speech, and conversational agents.
Models for what's available, accuracy, and domain customization.
API Reference for the classes, options, and C API.

Support

Discord for live support.
GitHub Issues for bugs and feature requests.

License

Licensed under the MIT License. The models are MIT by default too, in every language and at every size — the only exceptions are the legacy non-streaming models for languages other than English, which stay under the non-commercial Moonshine Community License and are enumerated in LICENSE.

View on GitHub

Recent activity

commits and pull requests

Fixed use-after-free race between free_stream and transcribe_streamOpen
#227 by fra-shipper · updated 1mo ago
Fix use-after-free race in transcriber_map accessorsClosed
#224 by fra-shipper · updated 1mo ago
docs: clarify that MOONSHINE_MODEL_ARCH_BASE_STREAMING is not currently in model catalogMerged
#215 by ghchinoy · updated 1mo ago
Unmap the streaming decoder model files on destructionClosed
#217 by aconchillo · updated 1mo ago
P
Build, run, and drop each Docker image so the release does not lose amd64 tags.
Pete Warden committed 234f60f · 1mo ago
P
Skip on-device iOS latency in the release when Xcode has no Apple ID.
Pete Warden committed dca34c2 · 1mo ago
P
Keep test-assets out of the Swift xcframework zip.
Pete Warden committed 74a4a89 · 1mo ago
P
Only build the moonshine target for Swift so the release doesn't fill the disk.
Pete Warden committed 530c315 · 1mo ago
P
Stop mentioning 0.1.5 in the model catalog comments.
Pete Warden committed d71fc16 · 1mo ago
P
Added testing and evaluation for the new language models
Pete Warden committed 76817fe · 1mo ago
P
License: make MIT the default for all models, with an enumerated non-commercial exception list
Pete Warden committed 547d006 · 1mo ago
P
Pin HF version for asset fetching (#221) and code formatting
Pete Warden committed 77ee8a8 · 1mo ago

Recent open issues

view all

Is everything under micro/ MIT licensed?
#228 · 16d ago · by mikeysklar · 1 comments
MicTranscriber: optional live audio-level callback across SDKs (Android use case)
#226 · 1mo ago · by sanogueralorenzo
Android: process crashes (SIGABRT "destroyed mutex") when close() is called while transcribing
#223 · 1mo ago · by monneyboi · 1 comments
Voice cloning turns American
#220 · 1mo ago · by mikey0000 · 3 comments
German dict uses ʁ instead of r
#213 · 2mo ago · by cryptomilk
German dict stresses the onset, not the vowel
#212 · 2mo ago · by cryptomilk · 4 comments

Discussions

all 15

Italian laguage model
Q&A · 1mo ago
OpenxAI Labs x Moonshine
General · 4mo ago
Are there any plans to add the Russian language?
Q&A · 4mo ago
Testing request: React Native Moonshine wrapper
Show and tell · 4mo ago
Link Problem
Q&A · 5mo ago

Releases and announcements

23 total

v0.1.5v0.1.5Aug 24, 20264.3K downloads
Release v0.1.5
v0.1.3v0.1.3Aug 18, 20262.5K downloads
Release v0.1.3
v0.1.2v0.1.2Aug 13, 20263.9K downloads
Release v0.1.2
v0.1.1v0.1.1Aug 7, 20266.9K downloads
Release v0.1.1
v0.1.0v0.1.0Jul 27, 202613.4K downloads
Release v0.1.0

Code frequency

additions and deletions

+3.2M lines added, -2.2M removed over the last year.

Commits per week

last 52 weeks

660 commits in the last 52 weeks.

When work happens

weekday and hour

Commit volume by weekday and hour (UTC). Larger dots mean more commits.

Date	List	Rank	Stars gained
Jul 21, 2026	daily	#13	+4
Jul 20, 2026	daily	#21	+3
Jul 19, 2026	daily	#8	+7
Feb 27, 2026	daily	#24	+154

tensorflow/tensorflow
An Open Source Machine Learning Framework for Everyone
200.7K stars · C++
ggml-org/llama.cpp
LLM inference in C/C++
130.3K stars · C++
unslothai/unsloth
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.
77.2K stars · Python
ocornut/imgui
Dear ImGui: Bloat-free Graphical User interface for C++ with minimal dependencies
76.5K stars · C++
protocolbuffers/protobuf
Protocol Buffers - Google's data interchange format
72.1K stars · C++
LadybirdBrowser/ladybird
Truly independent web browser
66.4K stars · C++