Google's new Gemini 3.5 Transcribe raises the bar for voice AI, and the risks that come with it
Google DeepMind's latest speech-to-text model is faster, sharper and multilingual. It also gives scammers a cleaner tool to work with.
Key points
- Google DeepMind launched Gemini 3.5 Transcribe on the day of announcement, calling it its most accurate speech-to-text model yet.
- The model scores a 4.0% Word Error Rate in streaming mode and 2.6% for pre-recorded audio, as measured by Artificial Analysis.
- Time to final transcription is 70% faster than Google's previous Chirp 3 model, again per Artificial Analysis.
- The system supports over 85 languages, identifies up to three speakers, and can be tuned to custom vocabulary.
- It is in public preview through the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform.
Google DeepMind has released a new speech-to-text engine called Gemini 3.5 Transcribe. Speech-to-text is the technology that turns spoken audio into written words, the thing your phone does when you dictate a message.
The pitch is simple. It listens more accurately, in more languages, with less delay, and cleans up your "ums" and false starts on the way to the page.
For a nurse dictating notes or a small business owner running a call centre, that sounds like pure upside. My job on the security beat is to tell you where the sharp edges are.
What is actually new here?
Google DeepMind says Gemini 3.5 Transcribe is its most precise transcription model to date. It comes in two flavours: a live streaming version for real-time apps, and a batch version for recorded audio like meetings and call logs.
The numbers matter. On a benchmark run by the independent firm Artificial Analysis, the model hits a 4.0% Word Error Rate in streaming and 2.6% for pre-recorded audio. Word Error Rate is the share of words the system gets wrong, so lower is better.
Speed also jumped. Time to final transcription improved by 70% compared with Chirp 3, Google's earlier model.
| Feature | Streaming | Pre-recorded |
|---|---|---|
| Word Error Rate (Artificial Analysis) | 4.0% | 2.6% |
| FLEURS multilingual WER | 5.50% | 5.04% |
| Languages supported | 85+ | 85+ |
| Speaker attribution | No | Up to 3 |
The model also does "smart" cleanup. If you say "let's meet Tuesday, no, Wednesday," it writes Wednesday. It strips filler words and formats the text on the fly.
Where does this show up for ordinary people?
In a few places you probably already use. On Android, the Gboard keyboard gets a feature called Rambler that turns rambling speech into tidy text. The Gemini app on macOS gets voice commands that can pull in what's on your screen. Chrome will soon let you talk to type in any web field.
Developers can plug it into their own products through the Gemini API in Google AI Studio, and businesses can access it through the Gemini Enterprise Agent Platform.
Should users be worried?
Only a little, and mostly in a specific way. Better transcription is genuinely useful. It also lowers the cost of scams that rely on clean, believable audio.
Here is the honest picture. Voice phishing, sometimes called vishing, already uses AI-generated voices to impersonate a boss, a bank, or a relative in trouble. A transcription model that handles noisy calls, regional accents, and 85+ languages makes it easier to build fully automated scam call centres that can react to what a victim says in real time, in the victim's own language.
The model itself is not the attack. The attack is what people build on top of accurate, cheap, real-time voice AI. Google's own guardrails apply inside its products, but any developer with API access is building their own experience.
There is a second concern worth flagging. The macOS version can read your screen to improve accuracy. That is a permission, not a default, and you should treat it like you'd treat screen-sharing on a work call.
What readers should watch for
- A caller who sounds unusually smooth and responsive, but pushes urgency. Real people pause. Automated voice agents rarely do, unless they are trained to.
- Unexpected calls in your own language from services that normally use another. Cheap multilingual AI is why.
- Any app that asks for microphone plus screen access together. Grant one, not both, unless you trust the vendor.
- Voicemails that sound like a family member in trouble. Hang up, call the person back on a known number.
Credit where it's due: Artificial Analysis ran the benchmarks Google is citing, which is worth knowing when you read the marketing.
