Speech-to-Text On-Prem

Speech-to-Text On-Prem -- Google Cloud’s first hybrid AI product – allows for easy integration of Google speech-recognition technologies in users’ on-premises solutions.

It gives users complete control – from within their own data centers -- of speech data protected by data residency and compliance requirements.

The machine learning technology leverages Google-developed speech recognition models that the cloud provider says are more accurate and smaller in size, and require fewer computing resources to run, than other solutions. The application can be deployed as a container to any Google Kubernetes Engine (GKE) or Anthos cluster.

Key capabilities include high-quality transcription using Google’s deep learning neural network algorithms. Speech-to-Text On-Prem can be deployed efficiently with models that are less than 1 GB in size and consume minimal resources, and it’s fully compatible with Google Cloud’s Speech-to-Text API and client libraries. Customers can use Google Cloud’s pre-built Istio objects to seamlessly scale up to thousands of connections. The application also is integrated with Operations (formerly Stackdriver), and metadata logs can be exported to one centralized location.

The new offering currently supports five languages: English, Cantonese, French, Japanese and Spanish.

It’s generally available on Anthos via the Google Cloud Marketplace.