Add voice profiles

[aaron-gh]

A voice profile is a set of speech settings that you can switch to at once: a speech engine, language, voice, volume, rate and pitch, and Send long text a sentence at a time. For example, one profile can use a slower rate for long text while the usual settings stay fast. Profiles are added in Text-to-speech > Voice profiles. When a language is chosen, the voice list shows only the engine's voices in that language.

Backtalk default is always there, and is the existing text-to-speech settings, so nothing changes until you add a profile and switch to it.

Switching

• The Voice profile reading control replaces Speech engine, and keeps its setting, so it is in the reading controls for anyone who had added Speech engine. It goes through Backtalk default and then the profiles in the order of the Voice profiles screen. With no profiles, it says "No voice profiles".
• Voice profile in use on the Voice profiles screen.
• A Voice profile item in the Backtalk menu, off by default and shown only once a profile exists. It lists the profiles and says which is in use.
• Choose voice profile, Previous voice profile and Next voice profile actions for gestures and keyboard shortcuts. All are unassigned by default.

Behaviour

While a profile is in use, changing the speech rate or pitch with gestures or the reading controls changes that profile only. All speech uses the profile's engine and voice, so Backtalk doesn't switch language or dialect, and the Spoken language reading control and menu item are hidden. If Spoken language is still the selected reading control, swiping up or down says "Spoken language unavailable while using a voice profile".

The Voice profiles screen can reorder profiles by dragging, or with Move up and Move down actions, which say which profile it moved above or below. Rename and Delete are actions too, and a long press shows them. The Text-to-speech and Voice profiles screens show which profile is in use, and update when it is switched by a gesture, keyboard shortcut or the Backtalk menu.

A profile's name can't be blank, can't match another profile's name or Backtalk default (ignoring case and spaces), and is cut at 100 characters. If a damaged setting, such as one restored from a backup, leaves a blank or duplicate name or a profile listed twice, the profile gets the first unused "Voice profile N" name and is listed once. Other damaged settings give their defaults instead of crashing, and rate, pitch and volume are kept to the ranges that can be set.

Switching to a profile on another engine, then back before that engine has started, stays on the engine in use. If a profile's voice can't be set, its language is used, then the engine's default voice, so it never keeps another profile's voice. A voice that isn't installed is looked for once per engine, not on every utterance. Changing a profile's engine saves the engine, language and voice together, so the profile switches engines once.

FailoverTextToSpeech reads the profile in its own preference listener, like the language switches, so speaking doesn't read settings.

Limitations

• No language or dialect switching under a profile.
• Announce keyboard keys faster doesn't prepare keys while a profile is in use, because prepared speech is kept by language, not by voice. Speech prepared before switching is stopped, and keys are prepared again once Backtalk default is back.
• Braille display commands can't switch profiles: they are fixed chords, not mapped by the user.

Parts

• VoiceProfiles (utils): storage, names, order and the profile in use.
• VoiceProfileNames (talkback): names with Backtalk default's and numbered names from resources, used by the service, reading control and menu.
• VoiceProfilesFragment and VoiceProfileFragment: the list and one profile's settings.
• FailoverTextToSpeech: engine, voice, language and sentence sending from the profile in use; prepared speech left alone under a profile; a pending engine switch cancelled when switching back.
• SelectorController, SpeechRateAndPitchActor, LanguageActor: the reading control, rate and pitch per profile, and hiding Spoken language.
• VoiceProfileMenuProcessor, gesture and keyboard shortcut actions.
• The new strings have tools:ignore="MissingTranslation", so that CI lint passes before they are translated. The two Czech strings for Speech engine are removed, since the English ones are gone.
• differences.md has a Voice profiles section in place of Speech engine reading control.

Testing

• The new VoiceProfilesTest covers:
  • storage and order;
  • ids that are never reused;
  • deleting the profile in use;
  • wrapping through Backtalk default;
  • damaged settings, duplicate ids, and blank, duplicate and long names;
  • which settings count as profile settings;
  • that a profile never switches language.
• Tested on a Galaxy Z Fold7, and on an Android 16 emulator with Google's engine and a second engine:
  • "No voice profiles" with no profiles;
  • switching by gesture with the settings screens open;
  • switching away from an engine and back before the other started;
  • changing the engine of the profile in use;
  • choosing a voice;
  • Announce keyboard keys faster with a profile in use, through keyboard windows and screen off and on;
  • menu order, moving, renaming and deleting profiles, and restarting Backtalk.