← Back to live feed1 story
Developers turned EmbeddingGemma 2 into an on-device speech-to-text model. It can process an hour of audio in about 13 seconds on a base MacBook Air.
Sign in to suggest
Key sources
- DISCUSSIONamrrsnews.ycombinator.com
- SOURCE@tslshahir“transcribes an hour of audio in about 13 seconds: 288× real time”x.com
- SOURCE@tomaarsen“train it on 3,000hrs of Vietnamese audio. IT REACHED SOTA”x.com
- SOURCE@sundarpichai“outperforms some specialist models more than twice its size”x.com
- SUPPORT@tomaarsen“The 740M footprint is modular: text: 270M vision: 170M audio: 300M”x.com
- SUPPORT@ivanfioravanti“Google Embedding Gemma 2 is the DeepSeek moment of embedding models”x.com
- SUPPORT@fdaudens“A dev built DigUp, which is basically what search on your Mac should be”x.com
- SOURCEhuggingnewshuggingnews.com