Google launches Gemini 3.5 Transcribe for live and recorded speech
The dedicated model is designed to handle both real-time audio and prerecorded material across speech-heavy workflows.
The story
Google has introduced Gemini 3.5 Transcribe, a speech model designed for real-time and prerecorded audio applications.
Reliable transcription supports meetings, media, accessibility and customer service, but accuracy can vary with accents, background noise and specialized vocabulary. Latency and privacy are equally important in live settings.
Developers will look for transparent evaluations across languages and noisy environments, plus controls for data retention, speaker identification and sensitive recordings.
INNOVOX analysis
Reliable transcription supports meetings, media, accessibility and customer service, but accuracy can vary with accents, background noise and specialized vocabulary. Latency and privacy are equally important in live settings.
What to watch
Developers will look for transparent evaluations across languages and noisy environments, plus controls for data retention, speaker identification and sensitive recordings.