Google's Gemini 3.5 Transcribe: New AI Model for Audio Transcription Supports 85 Languages and Removes 'Ums' and 'Ahs'

It's a new Google model dedicated to audio transcription, supporting over 85 languages, automatically removing speech disfluencies, and understanding specialized jargon, making it ideal for meetings and audio content.

It offers higher accuracy in noisy environments, better understanding of informal speech, and broader language support, compared to earlier models that struggled with background noise or interrupted speech.

Yes, especially with Arabic and many regional languages supported, enterprises can automate meeting transcription, improve voice customer service, and create accurate text content from audio.

Google promised a June release, but no specific date has been announced yet, leaving users waiting for the more advanced model.