Google’s AI Now Cleans Up Your Mumbling

Google just rolled out a smarter way to turn your spoken words into clean, readable text. The new tool, called **Gemini 3.5 Transcribe**, listens to your voice and automatically cuts out filler words like ‘um’ or ‘uh,’ so your transcripts sound polished without extra editing. It’s not just about fluff—it also recognizes over 85 languages, adapts to specialized jargon, and formats the text for you, saving you the hassle of cleaning up recordings manually.

This isn’t Google’s first rodeo with AI transcription, but it’s a big leap forward. The company claims this model does a better job than its older tool, **Chirp 3**, especially when it comes to handling multiple languages and reducing errors. For example, if you’re recording a meeting with colleagues who speak different languages, the AI can keep up and transcribe accurately. You can even teach it to spell words or terms your way—like brand names or technical terms—so it doesn’t “correct” them to something wrong.

Right now, the tool is available for Mac users of the Gemini app and Android users in select countries. Google says Chrome support is coming soon, and developers can try it out in their apps. The goal? To make transcription as effortless as speaking, whether you’re dictating notes, transcribing an interview, or just trying to sound more professional in a voice memo.

There’s one catch: Google also teased two other AI tools (**Gemini 3.5 Live** and **3.5 Live Experimental**) that handle live conversations and real-time reasoning, but it later clarified they aren’t launching today. So for now, the focus is on transcription—no live fixes, no mid-sentence interruptions handled automatically. But if you’ve ever waded through a recording full of ‘ums,’ this could be a game-changer.