By Geoff Adams-Spink BBC News Online disability affairs reporter

The AudioRead does away with robotic voices

The development is a result of a partnership between the Royal National Institute of the Blind (RNIB) and a company called Rhetorical Systems.

Rhetorical's rVoice text-to-speech technology (TTS) is based on the voice files of real people and sounds which are considerably more natural than traditional speech synthesisers.

Using robotic-sounding speech synthesis to read long newspaper or magazine articles is hard on the ear.

But having a real person read an entire newspaper on tape takes time and involves considerable cost.

Skip and jump

The RNIB plans to harness Rhetorical's TTS technology to deliver information to subscribers via a device called the AudioRead.

It is like a portable MP3 player that will dial in to retrieve information which people can listen to whenever and wherever they like.

The mainstay of our users who listen to books have never liked the idea of synthetic speech, they've always wanted a real reader Steve Tyler, RNIB

Unlike tape, users will also be able to navigate the newspaper, skipping between articles and sections at will.

Speaking to BBC News Online at the recent RNIB Techshare conference in Birmingham, the organisation's technology expert, Steve Tyler, said the organisation was "very excited" about the technology.

"We'll be able to produce 400 hours of audio in one hour," he said.

"It's one of the most human-like voices we've come across.

"The mainstay of our users who listen to books have never liked the idea of synthetic speech, they've always wanted a real reader."

Mr Tyler hopes that this solution will be acceptable to those people and trials of the AudioRead will start next month.

Pay to hear

The RNIB said one aspect they are keen to evaluate is whether subscribers are prepared to pay for such a wealth of information.

"People haven't had access to this kind of material before, ever," said Mr Tyler.

"At the same time they've never had to pay the costs."

People will be charged the annual equivalent of the cost of a daily paper or weekly or monthly magazine.

"The question is, will people do that? Is it sufficiently interesting," said Mr Tyler.

Rhetorical produces the synthetic voices by having someone read a standard script.

The various sounds are then "sliced and diced" and used in the company's speech engine.

A whole voice can be produced in around five weeks.