Google's new AI transcription edits out your 'ums' and 'ahs'
News Source
β’2026-08-26T13:00:00-04:00
π° What Happened
Google has updated Gemini Audio with new transcription capabilities that automatically detect specialised jargon and support more than 85 languages. The new model, called Gemini 3.5 Transcribe, is a fresh addition to the Gemini family. The Verge frames it as a system that can produce a cleaner transcript by editing out hesitations and filler speech. The announcement follows the launch of 3.5 Live Translate while users still wait for a long-overdue Gemini 3.5 Pro.
π The Backstory
The Gemini audio and translation line is Google's growing suite of speech-focused models layered on its larger LLM family. Detecting specialised terminology and rephrasing spoken filler out of transcription responds to the demand for usable, polished audio transcripts. The segment sits alongside translation and the wider race among AI labs to ship practical speech products, even as other parts of Google's model roadmap slip behind schedule.
π― Why It Matters
A transcript that automatically removes 'ums' and 'ahs' changes how speech is converted to text, useful for creators and journalists but raising questions to authenticity when AI rewrites what a speaker actually said.
Google has updated Gemini Audio with new transcription capabilities that automatically detect specialised jargon and support more than 85 languages. The new model, called Gemini 3.5 Transcribe, is a fresh addition to the Gemini family. The Verge frames it as a system that can produce a cleaner transcript by editing out hesitations and filler speech. The announcement follows the launch of 3.5 Live Translate while users still wait for a long-overdue Gemini 3.5 Pro.
The Gemini audio and translation line is Google's growing suite of speech-focused models layered on its larger LLM family. Detecting specialised terminology and rephrasing spoken filler out of transcription responds to the demand for usable, polished audio transcripts. The segment sits alongside translation and the wider race among AI labs to ship practical speech products, even as other parts of Google's model roadmap slip behind schedule.
A transcript that automatically removes 'ums' and 'ahs' changes how speech is converted to text, useful for creators and journalists but raising questions to authenticity when AI rewrites what a speaker actually said.