«

»

Print this Post

November 2016: Voice Banking update

Voice Banking

In the last 12 months, advances in voice banking have allowed more people to record their own voices to create an individualised synthetic voice for use in a communication device.

What is the difference between voice banking and phrase banking?

Some people like the idea of using their own voice when it becomes necessary to use a Speech Generating Device (SGD) to communicate.

Phrase banking is the recording of one’s speech/voice (whole words or phrases). Recorded messages can then be programmed into a SGD and used when needed for communication.

Voice Banking involves recording a sample of a person’s speech and creating a synthesised voice similar to the original recorded voice.  This voice can then be set as the ‘in-built’ voice within a SGD. Any message typed into the device can be spoken with the synthesised voice.

Creating a synthesised voice using Model Talker

The ModelTalker System is a speech synthesis software package developed by the Nemours Speech Research Laboratory and is designed to benefit people who are losing or who have already lost their ability to speak. https://www.modeltalker.org/

ModelTalker can be used to create a synthetic voice from voice banked phrases, therefore using the person’s own voice as the basis for the synthetic voice.  It is important to note that whilst this will sound similar to the person’s real voice, it is not an exact replica. It is still recommended to phrase bank important personal phrases through more traditional methods.

What is involved?

In order to create a synthesised voice, approximately 1600 phrases are recorded by the user over a period of several weeks.  This generally occurs in their own home following training and instruction from a TalkLink Speech Language Therapist.   Starting this process whilst speech is relatively unimpaired is preferred but not essential.

Voice banking for ModelTalker is a lengthy, process requiring a significant commitment from the person recording his/her speech with no guarantee of an optimum outcome. If the client is not prepared to commit their time despite this, then ModelTalker is likely not the best option for them.

What you will need

  • A laptop or computer running Windows 7 or above with very quiet fans. Noisy fans interfere with the recordings and make them unusable.
  • A broadband internet connection
  • A high quality USB headset.
  • A suitable environment to record in. TalkLink Tech Support can work with the client to help set this up. A suitable environment is one which has minimal echo, is small, has minimal background noise and lots of soft furnishings. A walk-in wardrobe is a great possibility. The room should be cool in temperature to avoid the computer’s fans from running at high speed and generating lots of noise. Fluorescent lighting which operates at 60Hz should be avoided as these create a buzz which is picked up on the recording. The environment must also be comfortable and be able to access the internet – the client will be spending a lot of time here!

To hear a ModelTalker synthesised voice created by a TalkLink client please see our YouTube Channel https://youtu.be/hBo79VATDqc.

For a PDF version of this handout, please download it here 201701-voice-banking-docx

Created by Jackie Hancock (SLT) and Kenny Ardouin (SLT)

The following two tabs change content below.
Paula Shennan

Paula Shennan

Speech Language Therapist at The Talklink Trust
I am a Speech and Language Therapist who joined the Talklink Auckland team in November 2011. I have an interest in the area of Assistive Technology and enabling and empowering individuals to communicate using various means.
Paula Shennan

Latest posts by Paula Shennan (see all)

Permanent link to this article: http://www.talklink.org.nz/index.php/2016/11/21/november-2016-voice-banking-update/

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>