Today for AI
HOT RADAR
toolsHEAT 18.0°

AI CLUSTERED EVENT · 10/13/2026

Google's EmbeddingGemma 2 Repurposed for Local Speech-to-Text: Transcribes 1 Hour in 13s on MacBook Air

1 reports archived1 independent sourcesupdated 10/13/2026, 19:27:14
Synthesis & Latest Updates
1 Sources Cross-Validated

Developers repurposed Google's multimodal embedding model EmbeddingGemma 2 into a high-speed local speech-to-text tool, transcribing one hour of audio in approximately 13 seconds on a base MacBook Air. The approach freezes the original audio encoder and trains only a lightweight 'reader' to map audio slices into the text space, running entirely on the Apple Neural Engine with no data leaving the device.

LATEST/Developers repurposed Google's multimodal embedding model EmbeddingGemma 2 into a high-speed local speech-to-text tool, transcribing one hour of audio in approximately 13 seconds on a base MacBook Air. The approach freezes the original audio encoder and trains only a lightweight 'reader' to map audio slices into the text space, running entirely on the Apple Neural Engine with no data leaving the device.

HEAT TRENDHourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. Hacker News AIT2·68 pts
    • Key innovation involves freezing EmbeddingGemma 2's audio encoder and training only a small 'reader' to map audio features into the shared text embedding space, avoiding full model retraining.
    • Extreme performance optimization: Achieves 288x real-time speed (1 hour of audio in 12.6 seconds) on a base M3 MacBook Air with 8GB RAM by leveraging the Apple Neural Engine.
    • Privacy and deployment advantages: The model is only 715 MB, runs fully offline, and supports global hotkey dictation, ideal for privacy-sensitive low-compute terminal scenarios.