What is custom speech?
Custom Speech in Azure allows users to enhance the accuracy of speech recognition for their applications and products by creating custom speech models. These models can be utilized for real-time speech to text, speech translation, and batch transcription. By training a custom model, users can improve recognition of domain-specific vocabulary and audio conditions unique to their application.
How does it work?
To leverage Custom Speech, users can upload their own data, test and train a custom model, compare accuracy between models, and deploy the model to a custom endpoint. The process involves creating a project, choosing a model, uploading test data, training the model with written transcripts and audio data, testing recognition quality, and finally deploying the model to a custom endpoint.
Choose your model
Custom Speech offers different approaches for using models. Users can opt for the base model, which provides accurate speech recognition for various scenarios and can be augmented with domain-specific vocabulary. Multiple custom models can be used for different areas within a domain with specific vocabularies. It is recommended to analyze the transcription from the base model and compare it with human-generated transcripts to determine if training a custom model is necessary.
Model stability and lifecycle
Once deployed to an endpoint, a base model or custom model in Custom Speech remains fixed until an update is decided. The accuracy and quality of speech recognition remain consistent even when a new base model is released. Users can utilize the model for a limited time, whether it is a trained custom model or a snapshot of a base model, until they choose to switch to a newer model.
Stay Ahead in Today’s Competitive Market!
Unlock your company’s full potential with a Virtual Delivery Center (VDC). Gain specialized expertise, drive
seamless operations, and scale effortlessly for long-term success.
Book a Meeting to Avail the Services of Azure Custom Speech Service