# Embedding model prototype claims hour-long transcription in about 13 seconds

_Published Tuesday, October 13, 2026 at 8:31 AM EDT · AI · Latest · Tier 2 — Notable_

![Embedding model prototype claims hour-long transcription in about 13 seconds — Primary](https://eg2.haku.sh/speed-table.png)

The developers of Embedding Gemma Transcribe say their speech-to-text prototype processes an hour of audio in about 13 seconds on a base MacBook Air, entirely on device. It keeps EmbeddingGemma 2's audio encoder unchanged and trains a small reader to map short audio slices into the model's text space, selecting the closest English word.

Only the reader was trained, using openly licensed English speech in under two hours on 8×H100. A separate tiny model adds punctuation and capitals. The developers report a word error rate of about 5%, against 2-3% for leading speech models, and caution that external speed benchmarks used different hardware and audio.

## Sources

- [eg2.haku.sh](https://eg2.haku.sh/)

---
Canonical: https://techandbusiness.org/newswire/41DaJg-fjj03aFCIb0ij9p
Published: 2026-10-13T12:31:02.094Z
Story chronology: 2026-10-13T11:27:14.000Z
Retrieved: 2026-10-13T14:49:51.915Z
Publisher: Tech & Business (techandbusiness.org)