Text to speech

This feature is only available on the following devices: Roku
Streaming Stick (3600X), Roku Express (3700X) and Express+ (3710X), Roku
Premiere (4620X) and Premiere+ (4630X), Roku Ultra (4640X), and any Roku
TV running Roku OS 7.2 and later.

Text to speech components

Components available since Roku OS 7.2

Text to speech (TTS) allows the developer to provide an audible spoken version of
the strings shown to the user in the app. For platforms that
are required to comply with the FCC Communications and Video Accessibility Act of 2010 (CVAA),
this capability can be used as part of compliance with CVAA, and the
current text to speech flite_tts library is built into the image. The
Roku text to speech capability supports different languages, voices,
rates of speech, volume of speech, and other aspects of text to speech.
Roku provides text to speech support in the following components,
interfaces, and events:

• roTextToSpeech
• ifTextToSpeech
• roTextToSpeechEvent

Components available since Roku OS 7.5

• roAudioGuide
• ifAudioGuide

Screen reader behavior for SceneGraph nodes

• ArrayGrid: speaks focused item (ContentMetaData::TITLE), followed by navigation hint, then ContentMetaData::AUDIO_GUIDE_SUFFIX (if any).

• Button: text of button is spoken only if focused

• ButtonGroup: speaks focused Button, followed
  by navigation hint (“button 1 of 4”), followed by button-specific hint, if any (Button-specific hint is spoken only for StarRatingButton).

• CheckList:
  speaks focused item (ContentMetaData::AUDIO_GUIDE_TEXT if any;
  otherwise
  ContentMetaData::TITLE)
  followed by navigation hint (“checkbox, checked, 1 of 4”)

• Dialog:
  speaks title, message, and bulletText (if any), then reads focused button

• Keyboard:
  speaks hint about caps lock toggling (once), then speaks focused key

• KeyboardDialog:
  speaks title, then keyboard

• Label:
  speaks text field

• LabelList:
  speaks focused ContentMetaData::AUDIO_GUIDE_TEXT if any; otherwise speaks
  ContentMetaData::TITLE, followed by navigation hint.

• MarkupGrid:
  speaks focused ContentMetaData::AUDIO_GUIDE_TEXT if any; otherwise speaks ContentMetaData::TITLE, followed by navigation hint, then
  ContentMetaData::AUDIO_GUIDE_SUFFIX (if any), then MEDIA speech (see below)

• MarkupList: speaks focused item
  (ContentMetaData::TITLE), followed by navigation hint, then ContentMetaData::AUDIO_GUIDE_SUFFIX (if any).

• MiniKeyboard:
  speaks focused key

• PinDialog:
  speaks dialog title, whether in key pad, then focused key or button

• PinPad:
  speaks focused key

• Poster: if focused, speaks audioGuideText field (if set)

• PosterGrid:
  speaks focused item
  (ContentMetaData::TITLE), followed by navigation hint.

• ProgressDialog:
  speaks dialog title, message, and bulletText every 15
  seconds. Speaks focused button if there is
  any.

• RadioButtonList:
  speaks focused item (ContentMetaData::AUDIO_GUIDE_TEXT if any;
  otherwise, ContentMetaData::TITLE), followed by navigation and
  selection
  hint

• RenderableNode:
  if speaking focused item (depends on context), will speak focused
  descendant; otherwise, will speak all descendants

• RowList:
  speaks row label (when row becomes focused), then speaks focused
  PosterGrid
  or
  MarkupGrid
  (MarkupGrid is used if itemComponentName is
  non-empty)

• ScrollableText:
  speaks text
  field

• ScrollingLabel:
  speaks text field

• Video: speaks
  HUD if displayed by user

Screen reader behavior for built-in SceneGraph panels and scenes:

• GridPanel:
  speaks panel, then leftLabel

• ListPanel:
  speaks panel, then leftLabel

• PanelSet:

  • If left panel is focused, speaks focused left panel, then
    unfocused right panel (if any)
  • If right panel is focused, speaks unfocused left panel, then
    focused right panel
  • If no panel is focused, speaks unfocused left panel, then
    unfocused right panel (if
    any)

• OverhangPanelSetScene:
  uses
  Overhang
  title when speaking location

• Scene: speaks
  dialog (if any); otherwise speaks
  PanelSet (if
  any); otherwise speaks as
  RenderableNode

MEDIA speech is spoken in the following order:

• ContentMetaData::TEXT
• ContentMetaData::DESCRIPTION
• ContentMetaData::DIRECTORS
• ContentMetaData::PRODUCERS
• ContentMetaData::ACTORS

There is no additional speech for the following nodes (they will behave the same as RenderableNode):

• BifDisplay
• BusySpinner
• LayoutGroup
• Overhang
• Panel
• ProgressBar
• Rectangle
• TextEditBox
• TrickPlayBar

Implementation tips

TTS interruptions

Many UI elements have default TTS behavior. It is possible that
speech triggered by these implementations can interrupt your TTS
implementation at times. You should keep track of the IDs of your TTS
utterances, as returned by say() and silence(), and handle interruptions
accordingly.

Other TTS implementation changes

Other TTS implementations may change the current voice, the current
language, the current volume, the current pitch, and/or the current
speech rate. You should keep track of how these parameters might change.

Long text delays

A long text string to be spoken by TTS may have a noticeable delay
before starting the speech, at least for the first speech of the long
string. For long text strings, you can break up the text string so that
the first speech is a reasonably short sentence, followed by longer
sentences as needed. You should not break up the long text string into
individual words, as it will affect phrasing without improving the
perceived delay in any noticeable way.