Voice banking

Also called Voice preservation, Personal synthetic voice, Custom voice

Recording your own voice while you still can, so that a synthetic voice can be built that sounds like you, and not like the tablet.

What it is

Voice banking is recording a large number of sentences in your own voice, before it changes or goes, so that a company can build a synthetic voice from them. The result is a text-to-speech voice that says anything you type, in a voice that is recognisably yours. Not a recording; a voice.

It is different from message banking, which keeps whole phrases exactly as you said them. Many people do both, and they are not in competition.

Who it can help

People who know their speech is likely to go: motor neurone disease above all, and also some people with Parkinson's, multiple sclerosis, Huntington's disease, a head and neck cancer with a planned laryngectomy, or a progressive dysarthria. It should be raised at diagnosis, not when speech is already slurred, because the recordings need a clear voice.

It is not useful to someone whose speech has already changed a great deal, though some services can now build a voice from a few minutes of older recordings or home video. And it is a personal choice: some people find hearing "themselves" from a machine comforting and some find it painful. Neither is wrong.

How it works

The person records sentences into a computer or phone with a decent microphone in a quiet room: traditionally 1,500 or more, taking many hours over weeks; newer services build a usable voice from 50 to a few hundred sentences, or from as little as ten minutes of speech, at a cost in naturalness. The service then trains a voice model and delivers it as a voice file that installs on a Windows communication aid, an iPad or an Android device, depending on the provider.

Some providers are free, some charge a few hundred pounds, and the MND Association has funded voice banking for people with MND. Some voices can later be tuned for age or clarity. The voice is a data file: it can be backed up, and it should be, in two places.

Getting started

If speech is expected to change, start this month. Ask your speech and language therapist which service they recommend and whether the cost is covered; then set up a quiet room, a USB or headset microphone, and half an hour a day. Record when the voice is freshest, usually mornings.

At the same time, start message banking: the laugh, the "love you", the way you say your children's names. Those cannot be synthesised. And record a few minutes of natural talk on your phone anyway, whatever else you do; the newer services can work from it.

Where to get help

The MND Association is the UK authority on voice banking and can advise on services and funding. Your speech and language therapist (referring to speech and language therapy) will support the process, and the RCSLT publishes professional guidance. The specialised AAC services install the finished voice on a funded device.