speak 0.2.3

Text-to-speech for Flutter on iOS, macOS, Android and web: awaitable prompts over other audio, no native code to set up. An alternative to flutter_tts.

0.2.3

• Android: voices() waits up to ten seconds after the engine starts for its voice list, instead of three: slower devices took longer to load their voices, and the first list came back without them.
• An utterance waits at most a second for the voice list. Until the list is there, it is spoken in its language with the engine's own choice of voice; on Android, while the engine is still loading, in its current voice rather than failing.

0.2.2

• Android: voices() no longer comes back empty when it is the first thing asked of a new Speak. Google's engine reports itself started before it has loaded its voices, or lists them before any is installed; for up to three seconds after the start, such a list is asked for again until it has an installed voice and settles.
• iOS and macOS: the voice list is read off the main thread. While the system cannot answer (a device or simulator setting up its voices), voices() waits, but the app's UI no longer does. Only voices that can be used are listed: a Mac still setting up its voices could list one that failed to speak (voiceUnavailable).
• Text with <, > or & is said as written. On macOS, before iOS 27, and in browsers on macOS, the system read markup in plain text: from a < on, words were lost. On Android, a text starting with < could be read as SSML.
• No word progress is reported while paused. An engine may report words it has prepared but not yet played (the iOS simulator under load did); they are reported on resume(), in order, or dropped if the utterance is stopped or interrupted instead.

0.2.1

• CallKit is no longer linked (App Review rejects apps containing CallKit on the China storefront). It is loaded at run time on the first question about calls; Speak(ios: IosSpeakOptions(callKit: false)) never loads it, and calls are then caught only by the session's interruption notification.

0.2.0

• Calls and other interruptions (Siri, alarms, another app taking the audio): what is being said ends as SpeakOutcome.interrupted, and an utterance asked for during a call ends as the new SpeakOutcome.blocked; nothing is replayed after the call. Only current evidence blocks, never a remembered interruption: iOS asks CallKit whether a call is going on and observes the session's interruption notifications (and their iOS 27 replacements); Android watches its audio focus and the audio mode.
• A claim refused for another reason (Android 15 in the background without a foreground service, an iOS session that cannot be activated) is no call: the text is spoken without it.
• SpeakAudio.duringCalls: SpeakDuringCalls.skip (the default) or .speak.
• Breaking: a switch over SpeakOutcome needs a blocked case.
• README: running in the background, per platform; the exact cause of the macOS failure under flutter run, and the debug-entitlement workaround.

0.1.3

• A clearer package description.

0.1.2

• The lead-in is played only when the audio goes to Bluetooth or a car (SpeakAudio.leadInOnBluetoothOnly, on by default): through the phone's speaker or wired headphones speech starts at once. The route is read right after the audio is claimed, on iOS from the audio session, on Android from where the speech's attributes route (Android 13+) or the connected Bluetooth outputs.
• Includes 0.1.1, which was tagged but never published: the README's guide for moving from flutter_tts, call by call.

0.1.1

• README: a guide for moving from flutter_tts, call by call.

0.1.0

• First release: text-to-speech on iOS, macOS, Android and the web.
• Awaitable utterances with an outcome (completed, interrupted, expired); queue or interrupt; expireAfter; stop.
• Voices with BCP-47 language, quality, network and install state; the best voice chosen for a language; availability for a language.
• Rate, pitch and volume, per call or as defaults.
• Word progress, pause and resume on every platform (Android resumes from the last reported word).
• Audio sharing per platform (SpeakAudio): iOS audio session and Android audio focus claimed right before speaking and given back after a short linger, or held with holdAudio. Modes mix, duck, duckAndPauseSpoken (SpeakAudio.navigation: music lowered, podcasts paused), pauseSpoken and pause.
• A silent leadIn before the first utterance after the audio is claimed, so Bluetooth headsets do not swallow the first syllable.
• Android engine choice (Speak.androidEngines, AndroidSpeakOptions).
• Per-utterance timeouts, and a stop that never waits on a silent engine.