How to Subscribe to the Watson STT Service

Last updated: June 2024

IBM provides a number of cloud-based services, including the Watson speech-to-text (STT) service. To use an STT service, you use your microphone to capture a spoken command to an audio file, send the audio file to the STT service, the service translates the audio to text, and returns the translated text to you. OrbWiz performs all of these steps for you — you don’t need to worry about the details — but it first needs you to subscribe to the STT service.

Why does OrbWiz use a STT service, instead of doing the speech-to-text conversion itself?

And why does OrbWiz use IBM’s Watson STT service?

OrbWiz provides a framework for building customizable digital assistants. Speech-to-text conversion is one part of that framework. Many companies now provide STT services, and OrbWiz’ philosophy is that it does not make sense to reinvent the wheel. It is better to rely on an established, best-in-class STT service. STT is constantly evolving, and established providers can be relied on to keep pace with the latest developments, and provide cutting edge STT performance.

Watson STT service has technical advantages over some other STT services. It allows the user to specify custom vocabulary to be used in STT. This is advantageous if, for example, you work in a specialized field with its own technical terminology. Watson STT allows you to incorporate specialized terms or expressions into your STT requests.

Watson STT also allows the user to specify speech recognition grammars. A grammar allows the user to tell the STT service to give more weight to certain expressions. In a simple case, if your digital assistant expects a yes-or-no answer, you can supply the STT service with a yes-or-no grammar, that will encourage it to favor a yes-or-no translation. Despite evolving technology, STT services are not perfect, and can be prone to errors. OrbWiz allows users to define grammars, that are passed to the Watson STT service to help improve the accuracy of the speech-to-text conversion. More technical detail on grammars is given at Speech Recognition Grammars.

Subscribing to Watson STT

The following provides some guidance to get started with Watson STT. However, IBM’s documentation remains the best resource.

At the time or writing, Watson STT provides three levels of subscription: a free Lite plan, a Plus plan that costs $0.02USD per minute, and a Premium plan that offers encryption and other features. Because it does not support custom vocabulary or grammars, OrbWiz will not work with the Lite plan. Instructions below explain how you can subscribe to the Plus plan.

To enrol in the Watson STT Plus plan:

The API key enables use of the STT service, which is billed to you. Keep it somewhere private to avoid anyone else using the account at your expense.

You will need to enter the Watson STT URL and API Key into OrbWiz. OrbWiz gives you two options. You can enter them once, and opt to have OrbWiz save them in the cloud. In this case the URL and API Key are stored in encrypted form in an OrbWiz cloud database, and are loaded automatically each time you start OrbWiz. Alternatively, you can elect to enter the URL and API Key manually, or import them from file, each time you start OrbWiz.