Google Launches Gemini 3.5 Transcribe for Speech-to-Text Conversion

Google Launches Gemini 3.5 Transcribe: The New Leap in Voice Transcription

While many are waiting (perhaps in vain) for the launch of Gemini 3.5 Pro, Google surprises with another novelty in the 3.5 line. Meet Gemini 3.5 Transcribe, an AI model that promises to transform voice input into text more fluidly, eliminating those annoying pauses like "um" and "uh," and delivering a polished final text. This model is already behind the "Rambler" feature in the Gboard of the Pixel 11, but will soon be present throughout the Google ecosystem.

What makes Gemini 3.5 Transcribe stand out? According to Google, it is significantly faster and more accurate than its predecessor, Chirp 3. The new AI promises to be about 70% faster in converting voice to text, with a reduced error rate in live speech down to 5.5%. It may seem like a small improvement over Chirp 3's 7.32%, but any advancement is welcome when it comes to avoiding manual corrections.

The Magic Behind Gemini 3.5 Transcribe

What really catches attention in Gemini 3.5 Transcribe is its ability to understand not just what you say, but what you mean. It removes the natural hesitations of speech and can even correct the text in real-time if you self-correct during speaking. Additionally, it is capable of handling specialized jargon, thanks to a customized vocabulary. All of this in 85 languages and with support for up to three speakers in pre-recorded audio.

But, as with everything in life, there is a catch. You are trusting the AI to capture the essence of your speech. For short texts, the model seems effective in cleaning up inconsistencies and verbal stumbles, but it may alter what you said, which is not always ideal. Trusting AI to correctly interpret what we mean is a crucial point.

Where to Find Gemini 3.5 Transcribe

Gemini 3.5 Transcribe is already active in some places, such as in the Rambler of Gboard, although still limited to Pixel 11 phones. Google promises to expand Rambler to more devices with Gemini Intelligence later this year. Users of the Gemini app on macOS can also already take advantage of this technology.

For developers, Gemini 3.5 Transcribe is already available in Antigravity, with full access to screen context and chat history (with your permission, of course). The model is also integrated into AI Studio, allowing the creation of apps with voice-optimized code. And if you are not using these specific apps or devices, don't worry. Google plans to bring AI transcription to the Chrome browser "soon," allowing you to input text in any web field, whether for writing emails, leaving comments, or even triggering voice chatbots.

In practice, Gemini 3.5 Transcribe is not just a technical update. It represents a significant step in how we interact with technology, making communication more natural and less prone to errors. And this, in an increasingly digital world, makes all the difference.