Gemini 3.5 Transcribe Is Coming to Chrome, and You'll Be Able to Talk to Type in Any Web Field

In short: Google has launched Gemini 3.5 Transcribe, a new AI speech-to-text model that cleans up filler words and formats speech automatically. It already powers Gboard's Rambler feature and the Gemini app on macOS, and Google says it's coming soon to Chrome, letting users dictate into any web field.

Key Takeaways:

• Gemini 3.5 Transcribe is Google's most precise speech-to-text model yet, replacing the older Chirp 3 model.

• It recognizes more than 85 languages and cleans up filler words, self-corrections, and formatting automatically.

Table of Contents

• What Is Gemini 3.5 Transcribe?

• What Changes When It Comes to Chrome?

What is Gemini 3.5 Transcribe? It's Google's newest AI audio model built specifically for speech-to-text. Instead of just converting sound into text word-for-word, it interprets natural speech — cutting filler words, fixing self-corrections on the fly, and auto-formatting the result into clean, readable text.

Google introduced the model this week, calling it the company's most precise speech-to-text system to date, according to Google's official announcement. It's designed to replace Chirp 3, the transcription model Google shipped across its products in 2025.

What Changes When It Comes to Chrome?

The headline detail for everyday users is what's coming next. Google confirmed that Gemini 3.5 Transcribe will soon power voice typing inside Google Chrome, letting people talk instead of type in any text field on any website — not just Google's own apps. That means dictating an email reply, drafting a social media post, filling out a form, or even prompting an AI chatbot, all by speaking instead of typing, according to Engadget's report on the launch.

Google hasn't given an exact rollout date for the Chrome feature yet, only describing it as "coming soon." The model is also headed to Google Search Live, Gemini Live, Docs, Keep, and Gmail, extending the same dictation experience across the company's productivity apps.

For a browser used by billions of people daily, folding an AI transcription model directly into any input field is a meaningfully bigger move than adding it to a single Google app. It effectively turns voice dictation into a browser-level feature rather than something locked to specific software.

It also puts Google in more direct competition with dictation tools built into rival platforms and third-party apps that people already use for hands-free typing. Where those tools often require a separate download or extension, Chrome-level integration means the feature would simply be there, available the moment Google flips the switch — no install required.

How Accurate Is It, Really?

Google backed the launch with hard numbers rather than just marketing language. According to benchmarks Google cited from Artificial Analysis, the model achieves a word error rate of about 4% in streaming mode and 2.6% in non-streaming use, based on details reported by 9to5Google. Google also says the time it takes to produce a final transcription has improved by roughly 70% compared with Chirp 3.

MetricGemini 3.5 TranscribePrevious Model (Chirp 3)Word Error Rate (streaming)~4.0%Higher (unspecified by Google)Word Error Rate (non-streaming)~2.6%Higher (unspecified by Google)Transcription speed~70% fasterBaselineLanguages supported85+FewerMulti-speaker detectionUp to 3 speakers (timestamped)Not featured

The model can also identify and timestamp up to three separate speakers in a pre-recorded audio clip, with support for more than three speakers still labeled experimental. That makes it useful for things like transcribing a meeting recording or a podcast episode and knowing who said what, without manually labeling each line afterward.

On top of transcription, it supports function calling, meaning it can hand off tasks — like generating an image — to other Gemini models while you're speaking, without needing to stop and type a separate prompt. In practice, Google says that's already showing up in the Gemini app on macOS, where a spoken request can trigger a background task like summarizing a file or searching for information, all without touching the keyboard.

Where Else Is the Model Already Live?

Gemini 3.5 Transcribe isn't a Chrome-only story just yet — it's already running behind the scenes in a few places:

• Gboard's Rambler feature on Android phones, including the Pixel 11 series, where it turns rambling spoken thoughts into structured text.

• The Gemini app on macOS, where it pairs with other Gemini models to carry out multi-step tasks using just voice input.

• Google Antigravity, Google's agentic development platform, where developers can use voice to "vibe code" apps.

• The Gemini API, available now in public preview through Google AI Studio, so third-party developers can build their own voice tools on top of it.

This kind of AI-driven price and workflow pressure is already rippling through the broader tech industry — ByteBreaking recently reported on how rising AI demand is pushing up prices for everyday gadgets, and on how AI's economic impact is playing out unevenly across the US workforce.

The Chrome rollout is the piece that will put Gemini 3.5 Transcribe in front of the widest audience yet, since Chrome remains the world's most used web browser. Once it ships, voice typing stops being a feature you have to seek out in a specific app and becomes something available by default almost anywhere you're typing online. For now, users will have to wait for Google to confirm a firm rollout date before they can try it inside Chrome itself.

FAQs

What is Gemini 3.5 Transcribe? Gemini 3.5 Transcribe is Google's newest AI speech-to-text model. It converts spoken audio into clean, formatted text, automatically removing filler words and correcting self-corrections as you speak.

When is Gemini 3.5 Transcribe coming to Chrome? Google has not announced an exact date. The company has only said the Chrome voice-typing feature is "coming soon," alongside a rollout to Search Live, Gemini Live, Docs, Keep, and Gmail.

Is Gemini 3.5 Transcribe available to developers? Yes. Gemini 3.5 Transcribe is available in public preview through the Gemini API in Google AI Studio and Google Antigravity, so developers can build voice agents, captioning tools, or analytics pipelines with it.

How many languages does Gemini 3.5 Transcribe support? Gemini 3.5 Transcribe supports more than 85 languages, with recognition for regional accents and dialects.

Does Gemini 3.5 Transcribe replace Chirp 3? Yes. Google says Gemini 3.5 Transcribe replaces Chirp 3, its 2025 transcription model, offering better accuracy and roughly 70% faster transcription times.