Skip to content
Trending storyDeveloping

Paige Bailey says Gemma models can run speech-to-text on-device

1 article1 sourcesince Oct 12Last article 57m ago ·

Overview

AISummary of 1 article

Paige Bailey, a Google developer relations engineer, says Gemma 4 and EmbeddingGemma can do speech-to-text efficiently, and cites open models as letting the community extend them in unexpected ways.

In a linked post, developer Shahir reports turning EmbeddingGemma 2 into a speech-to-text model that transcribes an hour of audio in about 13 seconds on a base M3 MacBook Air, or 288 times real time. Bailey's post does not include benchmark figures, and the timing is Shahir's self-reported result, not independently verified.

Written by AI from the articles below · updated Oct 12, 4:20 PM ET

Check the sources:

Latest developmentOct 12, 4:17 PM ET
Gemma 4 and EmbeddingGemma can do efficient speech-to-text

Article timeline

The articles in this story. Times are ET.

Oct 12
  1. 👩‍💻 Paige BaileyX
    Gemma 4 and EmbeddingGemma can do efficient speech-to-text

    AIPaige Bailey says open models let the community extend them in unexpected ways, and points to Gemma 4 and EmbeddingGemma doing speech-to-text efficiently. A linked post by Shahir reports turning EmbeddingGemma 2 into a speech-to-text model that transcribes an hour of audio in about 13 seconds on a base M3 MacBook Air, or 288× real time.

Heat trend

Not enough continuous observations to show a trend yet.