Top AI Product

Moonshine is an open-source voice stack from Moonshine AI (formerly Useful Sensors), founded by Pete Warden — the engineer behind TensorFlow Lite at Google. Speech-to-text, intent recognition, and text-to-speech, all in C++, all on-device. The Micro build runs a complete voice interface in 470KB of RAM on an 80-cent Raspberry Pi RP2350 chip. No cloud, no API keys. 9,001 GitHub stars, 330 points on the Hacker News front page.

Six times smaller than Whisper, more accurate

Moonshine’s largest speech-to-text model has 245 million parameters and scores 6.65% word error rate on the OpenASR leaderboard. Whisper Large v3: 1.5 billion parameters, 7.44%. It also streams — transcribing while you’re still talking — so voice agents respond an order of magnitude faster than batch-mode alternatives.

The SDK: embed it and go

It ships as a C++ library plus a Swift package, with Python, Android, and web bindings, MIT licensed. Drop it into an app and you get local STT and TTS — voice agents on wearables, toys, and appliances, or privacy-first dictation, all offline. Everyone else stacks gigabytes of weights for voice; Moonshine runs on a microcontroller. That’s the scarce direction.

You Might Also Like

• Insforge Hits 1 on Product Hunt and 3600 Github Stars is This What Agent Native Backends Look Like
• Openviking Treats ai Agent Memory Like a File System and 9k Github Stars say its Working
• Alibabas Agentscope Hits 21k Github Stars What Makes This Multi Agent Framework Different
• Agent Reach Hits 12k Github Stars by Solving ai Agents Biggest Blind Spot
• Dexter Passed 24k Stars on Github an Autonomous Agent Doing Real Equity Research