Flint

Hansson & Larsson is developing a system, which can ”understand” a simple vocabulary and translate utterances (words, numbers or sentences) into commands. This system will be accessed as a service making it extremely easy to implement into an existing applications, be that a website or a desktop program. It will provide an open API on top of which others can create even more powerful services on their own.

Technical description

Flint is implemented as an application running inside the Red5 Flash Server, and incorporating state of the art speech recognition software. It is accessed by a Flash client application, running in a user's computer. The Flash client application is initialized with a URL specifying the Flint server, and a list of words and phrases. The Flash client application requests access to the user's microphone, and subsequently sends all audio to the Flint server, which analyzes the audio and executes a callback in the client, thus providing it with the matching word or phrase.

We will be providing a fully functioning Flash application, which in itself, together with some JavaScript, can be deployed on any website, thus making it speech controlled. We will also provide a small library containing the ActionScript source code to implement these features in any Flash application.

Open source

As we are strong supporters of open source, and we intend to release the project in full, as open source. This way, we hope to get others in the community involved in improving on the system's  functionality and features, and help spotting and fixing bugs.

Please check back in the near future for more information about the release.