Skip to content

SandFly 0.2.0

Latest

Choose a tag to compare

@github-actions github-actions released this 29 Oct 12:14
· 173 commits to master since this release

build: 822

New

  • An interactive tutorial teaches SandFly step by step, like VoiceOver's tutorial on the Mac: 16 lessons in 5 chapters, from exploring by touch to reading controls, typing, actions and the SandFly menu. You try each gesture, and SandFly checks it and says what happened. It opens the first time SandFly is turned on, and under Settings > Help & feedback > SandFly tutorial.
  • Settings > Commands lists every command a gesture or keyboard shortcut can run, in categories. Each command's screen lists its gestures, with Add gesture and Remove gesture, and its keyboard shortcut, which is recorded there. Every command can have gestures and a keyboard shortcut, including commands that only had one of them.
  • Translation. Translate last spoken phrase, a command, translates what SandFly said last and speaks the translation. Settings > Translation chooses the service (on-device, Google Cloud Translation, DeepL, Microsoft Translator, LibreTranslate or an AI service) and the languages. Automatic translation translates speech in another language before it is spoken, with a reading control to turn it on and off. Translate notifications translates notifications in another language, read in their voice profile's voice. On-device translation downloads each language when it is first needed, or all at once with Download all languages, and then works offline.
  • Arrow key navigation: with a keyboard, press Left and Right arrow together to turn it on or off. While it's on, the arrow keys run commands, as in VoiceOver on iOS: each arrow alone, two arrows together, and Control or Alt with an arrow. Change each one's command under Settings > Keyboard shortcuts.
  • Next track and Previous track commands control the media that is playing, as a headset's buttons do.
  • The Sound mode reading control sets the ringer to Sound, Vibrate or Mute. Add it under Settings > Reading controls.
  • Settings > Typing > Dictation > Auto stop after silence ends recording after a pause of 3 to 30 seconds or 1 minute. It is off by default.
  • New features in SandFly, at the top of the settings, shows what is new in this version. It also opens once after updating.
  • SandFly Pro is a paid add-on. Its AI agent operates apps for you, including apps and games that don't work with screen readers, and runs SandFly commands, such as changing the speech rate. Assign Start or stop AI agent task to a gesture or key under Settings > Commands, then say or type the task. A safety check asks before sensitive actions such as sending messages, and in banking, payment and similar apps the agent only reads. Sounds mark the start, end, questions, stops and failures of a task, and sound themes can replace them.
  • SandFly Pro is also an AI service for Describe image, Describe screen, live recognition, dictation correction and translation, with no API key of your own.
  • Buy a SandFly Pro code, enter it or add time under Settings > SandFly Pro. A code lasts a month or a year from when it is first entered and works on up to 3 devices.
  • SandFly Pro sends what it works on (screenshots, screen text, images, tasks and questions) to SandFly's server and Anthropic. It is not end-to-end encrypted. The privacy policy, under Settings > Help & feedback > Privacy policy, describes what is sent, kept and logged.

Changes

  • Settings are arranged by topic: every setting is on the screen named after what it changes, and the main screen lists those screens without group headings.
  • Reading controls and SandFly menu settings start with Current, the items in use in their order, each with Move up, Move down and Remove. Items not in use are listed below, and a double tap adds one at the end.
  • Every speech input uses SandFly's dictation, with the recognition and AI correction chosen in Dictation settings: Ask by voice after Describe image and Describe screen, and voice commands. They used Android's speech recognition.
  • Changing to a voice profile with another speech engine switches at once, and the profile's name is said in its own voice. The name was said by the old engine and cut off.
  • Changing voice profiles says the profile's name, not its engine's.
  • Speak text formatting moved to Verbosity > Text and symbols, and Keyboard echo to the top of Typing.
  • The voice command to open settings is "SandFly settings". "TalkBack settings" still works.
  • SandFly is translated into every language it supports, and the translations name settings as they appear on screen.
  • Android's backup no longer copies SandFly's settings, which hold API keys, to cloud backups. Use Export settings to move settings to another phone.
  • Settings > Help & feedback has Privacy policy.

Fixes

  • Speech stayed in the earpiece after a call during which Bluetooth earbuds were connected, on One UI 9.
  • The call length spoken when a call ended was cut off or not heard with Bluetooth earbuds.
  • After an update, the notification list stayed empty and notifications that only the notification listener sees weren't read, until SandFly was turned off and on.
  • A voice profile whose speech engine Android had stopped stayed silent.
  • Export settings wrote the API keys of Claude, OpenAI and other AI services, apart from Gemini, to the file.
  • In Chrome on Android 15 and earlier, checkboxes and radio buttons named only by their label were read as "Check box".
  • A web post that a site's keyboard shortcut moves to, such as on Mastodon, was read without its author, time and hashtags. White space between web elements was read as "Space".
  • The braille keyboard's submit gesture did nothing in Messages, WhatsApp, Telegram, Pachli and other apps that send with a button beside the message field. It now presses that button.