Enhanced Accuracy Through Multimodal Training
Google introduced Gemini 3.5 Transcribe, a new speech recognition system that converts spoken language into written text. The model already powers the latest version of Gboard's voice typing feature called Rambler. Google plans to bring the technology to Chrome browsers soon, though a specific timeline remains unclear.
Breaking news
Judge Questions Fairness of Google's AI Overviews in Antitrust Case
Apple's iPhone Event Tagline Varies by Country Ahead of September Launch
Gemini 3.5 Transcribe Eliminates Filler Words From Speech
Google addresses Keep list addition issues with potential fixGemini 3.5 Transcribe represents Google's latest advancement in artificial intelligence driven transcription. The system builds upon previous voice recognition capabilities integrated into Android keyboards and web browsers. Internal testing suggests improved accuracy in noisy environments and better handling of accents compared to earlier versions.
Google trained Gemini 3.5 Transcribe using diverse audio datasets including podcasts, phone calls, and real-world conversations. Engineers incorporated feedback from millions of daily voice interactions across Google services. Early benchmarks show roughly 15 percent fewer errors when transcribing informal speech compared to the prior model.
How Will This Affect Everyday Users?
The company emphasized privacy safeguards during development. Audio snippets processed through Gboard are anonymized before reaching Google's servers. Users can review and delete their voice history through account settings. Google claims the model performs well even when internet connectivity drops, thanks to on-device processing capabilities.
Most Android phone owners will notice changes immediately when updating Gboard. Voice typing now responds faster and handles longer phrases without cutting off mid-sentence. Chrome users on desktop and mobile can expect similar upgrades once Google rolls out browser-level integration later this year.
Developers gain access to Gemini 3.5 Transcribe through Google's ML Kit, allowing third-party apps to embed high-quality speech recognition. Early partners include language learning platforms and accessibility tools designed for hearing-impaired users.
Looking ahead, Google plans quarterly updates to refine performance based on user behavior patterns. The company did not confirm whether the model supports languages beyond English at launch. Analysts say this move strengthens Google's position against competitors like Apple and Microsoft in the voice AI space.
Frequently Asked Questions
Can I use Gemini 3.5 Transcribe offline? Yes, basic voice typing works without an internet connection on supported devices. Full accuracy improvements require cloud processing when available.
Is my voice data stored by Google? Audio recordings are anonymized and used only to improve the service unless you opt out through your Google Account settings.
Will Chrome get the new model automatically? Google will enable Gemini 3.5 Transcribe in Chrome through a gradual rollout starting later this year. Users may need to update their browser version.
