Gemini 3.5 Transcribe encourages you to let go of the pressure to be a flawless speaker.
Google
We all tend to ramble, backtrack, and trip over our words while speaking. Google’s new Gemini 3.5 Transcribe has been designed with our imperfect speech in mind. Heralded as the company's most accurate speech-to-text model to date, it takes your disorganized, raw speech and converts it into clean, formatted text without requiring you to sound like a news anchor first.
It launches alongside two additional models, Gemini 3.5 Live and Gemini 3.5 Live Experimental, collectively referred to as “Gemini Audio.”
We’re unveiling Gemini 3.5 Transcribe, our most accurate speech-to-text model yet 🔊 It converts audio into precise transcripts in over 85 languages, eliminating filler words like “ums” and “ahs,” while managing self-corrections and capturing your intent, enabling you to accomplish tasks using… pic.twitter.com/4GgHjSd301— Google (@Google) August 26, 2026
How Gemini 3.5 Transcribe enhances your speech
The model automatically removes filler words such as “um” and “ah,” ensuring that your transcribed text isn’t cluttered with verbal stumbles. It also recognizes real-time self-corrections, so if you say “let’s meet Tuesday, no wait, Wednesday,” it comprehends that you intended to say Wednesday and makes the necessary adjustment. You can also edit your text solely by using your voice.
Additionally, the model can assign tasks like image generation or file analysis to other Gemini models through function calling, a feature currently active in the Gemini app on macOS.
Gemini 3.5 Transcribe powers Rambler on @Android. We initially introduced Rambler at the Android Show to assist in transforming spoken thoughts into refined text. It utilizes 3.5 Transcribe to automatically eliminate filler words and refine speech. You can also make edits vocally, correcting… pic.twitter.com/BTHQuDYtkB— Google (@Google) August 26, 2026
Google reports impressive accuracy metrics to support this, with a word error rate of just 4% for real-time streaming and 2.6% for pre-recorded audio, plus it accommodates over 85 languages, regional accents, and specialized vocabulary. It can even attribute speech to up to three distinct speakers in a recording, complete with timestamps.
Which Google apps currently feature Gemini 3.5 Transcribe
Google
This model isn’t confined to a lab. It is already powering Rambler, the dictation feature on Android’s Gboard, and is also integrated into the Gemini app on macOS. Chrome support will be available soon, enabling you to dictate directly into any web field.
Developers can access it now through Google AI Studio and Antigravity, Google’s agentic coding platform. Enterprise users can use it via the Gemini Enterprise Agent Platform. It is also expanding into Search Live, Gemini Live, Docs, Keep, and Gmail. So, the next time you’re speaking into your phone, Gemini might just comprehend you better than you understand yourself.
Manisha Priyadarshini is a tech and entertainment writer with over nine years of editorial experience.
Claude Cowork receives its own browser that doesn’t interfere with your tabs, bookmarks, or saved passwords
Anthropic provides Claude with its own integrated browser that is distinct from your personal browser.
Claude Cowork has just introduced its own browser, built directly into the desktop app. Previously, granting Claude web access within Cowork meant providing it with your actual Chrome browser through the Claude in Chrome extension. Now there is an alternative. Claude can launch its own separate browser within the Cowork side panel, managing the web aspect of tasks independently while you continue to work on your current activities.
Read more
LibreOffice 26.8 introduces numerous features and avoids pushing AI onto users
No chatbots, no cloud sync, no subscription fees. Just meaningful enhancements for key office tasks.
While tech giants are in a relentless race to integrate generative AI into every facet of life, the team behind the open-source office suite LibreOffice is taking a different approach. The Document Foundation has released LibreOffice 26.8 for Windows, macOS, and Linux, focusing on fundamental, impactful enhancements for core office tasks without overcrowding your toolbars with unnecessary chatbots or diverting your documents to cloud services.
Practical features aimed at enhancing your workflow
Read more
AI receptionists appear to be a risky idea after a patient reported it failed to comprehend her request
Your GP utilizes an AI receptionist, and it hopes you articulate your words clearly.
AI receptionists are meant to simplify the process of contacting a doctor. However, an increasing number of patients are finding the opposite to be true, especially when a machine struggles to understand their manner of speaking, follow instructions, or convey their needs. A recent incident in South Yorkshire highlights this issue. Judith Butterfield, a 71-year-old woman recovering from
Другие статьи
Increasing expenses associated with the token consumption competition.
The reasons behind tokenmaxxing increasing AI expenses, the impact of model selection and context windows on LLM costs, and the importance for companies to prioritize results.
I tuned in to Gamescom 2026 so you don't have to — here are 6 games that are worth the wait.
Gamescom 2026 offered a plethora of trailers and announcements, but these six titles — ranging from Metro 2039 to Nodus Fall — are the ones that have stayed with me.
Salesforce is placing Claude at the core of its products and incorporating itself within Claude.
Salesforce and Anthropic have unveiled Claudeforce, designating Claude as the primary model for Agentforce and Slack, while incorporating Salesforce as a plugin within Claude.
Nvidia announced a revenue of $96.2 billion for the second quarter.
Nvidia reported a revenue of $96.2 billion and provided its first forecast for the upcoming year, with the CFO specifically discussing circular financing during the call.
The regulator in South Korea has stated that the restrictions imposed by Meta for teenagers should be enforced globally, rather than being limited to the United States.
South Korea's media regulator has contended that Meta's recently announced youth protections should be implemented globally, shortly after the company stated that these measures do not automatically apply everywhere.
Surgeons in London have performed the first-ever brain tumor removal surgery with the assistance of AI.
Surgeons at the National Hospital for Neurology and Neurosurgery in London successfully extracted an 11mm pituitary tumor using an AI system that analyzed the live feed from the endoscope.
Gemini 3.5 Transcribe encourages you to let go of the pressure to be a flawless speaker.
Google's Gemini 3.5 Transcribe automatically eliminates filler words such as "um" and "ah" from your spoken content, converting disjointed speech into polished, organized text.