Voice recognition has been a dream of many for the last 10 years. It's an illusive goal because interpreting speech is very complicated and takes a lot of computing power. Past solutions used a "train the system" model for interpretation. Users had to endure lengthy sessions of talking to the machine, before they can enjoy any real level of reliable speech recognition.

Wizzard Software has introduced the WizzScribe SI conversational voice recognition engine. It converts speech directly to text, independently of the speaker. It doesn't require any training of the system. The technology, pioneered by IBM, is aimed at developers that want to integrate speech recognition into their applications. Corporate users can deploy it internally to bring voice recognition, to their business processes.

The software, at $3400 per processor license, is designed for batch processing large volumes of audio speech in a server environment. Multiple processors, large amounts of memory, and a lot of disk space are the domain of this engine. Good thing it runs on Linux.

Next: Does it Work on a Consumer Machine? »