An hour of audio transcribed in about 13s on MacBook Air
Developers have repurposed Google's EmbeddingGemma 2 model to perform on-device speech-to-text transcription on MacBook Air hardware. The project achieves high speeds and decent accuracy without sending audio data to the cloud.
Why it matters
This demonstrates the growing capability of local, on-device AI models to perform complex tasks efficiently while maintaining user privacy.
Google's EmbeddingGemma 2 is an embedding model. We turned it into a speech-to-text model: an hour of audio in about 13 seconds on a base MacBook Air , entirely on device.
EmbeddingGemma 2 is a multimodal embedding model: text, images, audio and video in one shared space. Say a word out loud and its audio lands close to the written word.
So we tried reading words straight out of that space.
Only the reader is trained: under two hours on 8×H100, on openly licensed English speech. The base model was never trained for speech-to-text.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in