Google Releases Gemini 3.5 Transcribe Engine with Real-Time Context-Aware Speech AI
Published on August 28, 2026 · Source: Ars Technica
Google has launched Gemini 3.5 Transcribe, a specialized multimodal speech processing engine integrated into Gboard, Chrome, and Google Workspace applications.
Google today announced the general availability of Gemini 3.5 Transcribe, a high-throughput, low-latency speech recognition foundation model designed to run seamlessly across mobile edge chips and cloud infrastructure.
Built upon technology initially previewed in Gboard's experimental Rambler voice input, Gemini 3.5 Transcribe intelligently removes vocal hesitations ("ums" and "ahs"), formats domain-specific technical jargon, and inserts contextually accurate punctuation across more than 100 spoken languages simultaneously.
Developers can access the API endpoint for audio processing at a 75% reduction in compute cost compared to previous generation Whisper and Gemini 1.5 speech pipelines.