Need for Voices that Sound Like a Child is Speaking for TTS and Screenreaders

[SuzanneTaylor]

[This github entry is from the Accessibility for Children Community Group]

Children's voices can be helpful for authenticity, identification and engagement. It is particularly helpful in Instructional materials, Testing contexts, Games, and Children's Literature.

For example, when a story is written as though a child is sharing their experiences, children decoding Braille or printed text would likely imagine a child's voice reading, while a child who reads best through realistic sounding TTS might struggle to reassign the voice to a child in their mind, reducing engagement. When the narrator or speakers change back-and-forth from adults to children, it would be particularly difficult to accommodate this using device settings.

Another example is when software allows peer reviews or allows children to write text to each other. In this case, children's voices can provide a more realistic, relaxing and informal experience, while use of adult voices could be disengaging or even cause a misinterpretation of the authority level of the feedback.

Pronunciation can affect psychological responses, which are very important for learning. Drawing from the field of child psychology, "affective learning" (e.g. emotionally real for the student, emotionally engaging learning, learning from a relative) can be more effective than "instrumental learning" (e.g. from books or a traditional classroom experiences.) Using a child's voice can make authored materials and software-based experiences more like "affective learning".

The Accessibility for Children Community Group suggests the following changes to the document:

• In Section 2.2's title, include Age
• In Section 2.2 mention the need for a child's voice specifically and the advantages as indicated above
• In Section 2.2, mention that a way to revert to the user's chosen voice easily is needed in case a user (a child) has an adverse emotional reaction to a particular voice or is simply unable to comprehend a particular voice. For example, if the voice sounds like someone the user knows who has recently passed away. The child, their guardian, or a educator might execute those options.

[AutoSponge]

I think the third bullet, "mention [that] a way to revert to the user's chosen voice" may not be part of our spec. The ability to ignore an author intent/suggestion for voice pack/settings is between the user and their TTS engine or AT, not the user and the markup. That said, we can provide an aside in our document to note that this is an important consideration for TTS implementations.

[AutoSponge]

These other suggestions will be incorporated into our spec. Thanks! :)

[learningactivist]

In our system, we enable the content author to specify a default voice (Woman, Man, Girl, Boy) and then allow the parent or teacher to enable or disable whether the child can change the voice.

https://mlc.learningstewards.org/help/#Settings