Learn More About ModelTalker

Learn More About ModelTalker

  • by Therapy Box
  • Mar 11, 2019
  • 4 min read

ModelTalker is an engine for creating voices from recordings of a patient's natural voice. The result is a synthesized voice similar to what the person might sound like if he or she weren't limited by speech disorders. One of the great benefits of ModelTalker is that the synthetic voices created are virtually unlimited and can be used in the ModelTalker Text-to-Speech system to express words and phrases that have not been previously recorded by the user.

The ModelTalker System was developed by the Nemours Speech Research Laboratory located at the Alfred I. duPont Hospital for Children with funding from the National Institute for Disability and Rehabilitation Research, the National Institutes of Health, and Nemours Biomedical Research.

Who is it for?

The ModelTalker System is a revolutionary speech synthesis software that is specifically designed to benefit people who are losing or who have already lost their ability to speak. Potential benefits of this technology include those with ALS, MND, Dysarthria, or those with other speech disorders.

It allows these people to use a synthetic version of their own voice, or a voice best suited to represent their own voice. Voices can be built from recordings by potential/current AAC device users, the friends or family of an AAC device user, or clinicians or researchers.

How are we implementing it?

Your ModelTalker voice banked voice is downloadable directly from Predictable and ChatAble either via iTunes or via a direct download from the ModelTalker website. This means that users have immediate access to their synthesized voice, enabling them to communicate with others using their own voice!

How does it works?

  • Go to the ModelTalker website www.modeltalker.org.
  • Fill in and submit a ModelTalker Beta Software Registration form in order to receive a registration code and create a voice banking account.
  • Download the ModelTalker Voice Recorder (MTVR) or use the online recording tool.
  • In a quiet area, record the 10 screening sentences using a good quality head-mounted microphone.
  • ModelTalker will review your recordings to test their quality. In some cases, you will be sent further suggestions for how you can create better quality recordings.
  • If the quality of the recordings has been approved, you will receive the full inventory of approximately 1600 sentences. This should take about four to six hours to record, ideally distributed over three to four days.
  • Once all the phrases have been recorded and uploaded, ModelTalker will use them to build your synthetic voice.

How long will it take to record my voice?

This will depend entirely on the person who is recording the voice. Some people will be able to do this much faster than others. Usually, it will take approximately four to six hours over three to four days.

Will my synthetic voice sound just like me?

Your synthetic voice will definitely capture your natural voice rather well. However, be aware that because this is a synthetic voice, it will never sound exactly like you. It may sound slightly robotic in terms of timing and intonation. If you want to listen to some examples of synthetic voices and compare them to the natural voices from which they were built, please visit the ModelTalker website.

What microphone should I use to record my voice?

Ensure that you are not using a desktop microphone, a built-in laptop microphone or an analogue headset, as these can cause problems and low-quality recordings. Ideally, you should be using a head-mounted USB microphone to ensure that the quality of your recordings is as good as possible. The better your recordings, the better your synthetic voice will be. If you have any questions regarding recording, please contact ModelTalker directly.

How long will it take for me to receive my synthetic voice?

This should not take longer than a couple of weeks.