Can an embedding model transcribe speech?

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.

How it works

1. EmbeddingGemma 2's audio encoder stays exactly as it is.
2. A small reader moves each short slice of audio into the model's text space.
3. The closest of ~229,000 English words wins, and the slices become a transcript.
4. A tiny extra model adds punctuation and capitals.

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.

How fast?

× real time: seconds of audio per second of processing. Rows 1–2: our tests on the same base M3 MacBook Air. Rows 3–5: FluidAudio's published benchmarks on larger, newer Macs and different audio. A rough guide, not a like-for-like ranking.

Where it stands

Its word error rate is about 5% right now, against 2–3% for leading speech models. We believe a better-trained reader can close that gap; that's the next iteration.

Use it

Drop in an audio file, or hold Right Option, speak and release to dictate into any app. It runs on the Mac's Neural Engine; your audio never leaves the machine.