(a) Call stop() when we've written less than the
AudioTrack buffer. This forces pending buffers to
get mixed.
(b) Introduce a minimum sleep time to avoid spinlocks
if they system is busy
Change-Id: If70937e8b4e8c5d02d7dadc0d3086f97a10eb7ef
Also fixes a bug where PlaybackSynthesisCallback would
continue to pump audio to the audio playback thread after
being stopped.
Change-Id: I3233eb4858245b6e7c7df72e241a0708c668f1b4
Measurements have shown that the audio track in streaming mode
is pretty efficient as is, thanks to dalvik's JNI implementation.
Also, in streaming mode write( ) blocks until all data has been
copied to the mixer, which is the kind of behaviour we want so
that we don't have to explicitly block in our code. Also,
the java AudioTrack is thread-safe with respect to stop( ) calls
during write, which means we can do everything completeAudioAvailable
did, just as efficiently without it.
Change-Id: I70a8a108d94c7260676673979b5ea29f37fa5db4
(a) Fix an NPE reported when no TTS engine is installed.
(b) Fix a bad link, and some javadoc typos.
bug: 5004015
Change-Id: I38b97dbb5d7d4065a5ee408fae9138638ed48f40
Engines must declare a <meta-data> attribute in their
manifest with name "android.speech.tts". This must reference
an XML resource as per
com.android.internal.R.styleable.TextToSpeechEngine.
Change-Id: I56a6b9f1a360174f98c9f39da901ade83d0f38a7
This has been discussed in an email thread. This brings the
text to speech API in line with other android APIs.
The user can set default speech rates / languages and
engine preferences, which the API obeys BUT these can be
overriden by individual apps (much like intent.setComponent
and intent.setClassName can be used to launch a specific
handler for a given intent).
Also, all installed engines are enabled by default. The
user is shown the "data privacy" warning every time an
installed engine is set to the default.
Change-Id: I24f5f331b3a7cc7ce1a62962192e3a452bdca9d4
This makes it a lot easier for engine implementers
to test their engine. This is a dumb data object, and
contains no logic whatsoever.
Change-Id: I2a8bea134a8ff6b962508cfed67ea6085a682b23
(a) ensure that onInit is called exactly once per call to initTts().
(b) TtsEngines should never return an engine that is not installed.
Change-Id: Ic1ef63a9339e6d1050e3302445ca40d3ae481f61
One point to note is that no TTS engine will be selected
by default if there are no system TTS engines.
Change-Id: I397061e3cf29b8a1c66e7864262b43621ef5100d
(a) Document the undocumented queue mode (mode == 2)
(b) intern strings to fix a bug in removeCallbacksAndMessages
(c) remove an unnecessarly lock object, change the field it
guarded to volatile.
Change-Id: If46bbfe2afb047b3ba5b3cb9a90c292656379ec1
Note that this change is slightly more flexible, it defines
a priority for each playback request. We needn't really throw
away synthesis requests from other apps. We could just play
them after higher priority ones (for e.g from talkback).
Change-Id: I6c7dd9abe4871e87da08d9aaa4c55225a69078e5
This ended up making the implementation a lot cleaner
as well. See the bug listed below for some background.
bug:4553470
Change-Id: If16476a57e389c2f9b228f6548e426642d292b49
There is no point loading a language that will never be used.
It is in fact harmful, because the engine might load it and
then receive a request for the default voice and have to load
the default voice again.
Change-Id: I8082f28d15b6e325d6a39d439b66753711a2c244
Now, SynthesisRequest.audioAvailable( ) will not block
until all data has been written to the audiotrack.
Change-Id: Ie24c0b6873ae44659e88646fb7ef821750dc314e
This is a new API for writing text-to-speech engines.
The existing API for apps that use TTS remains the same,
with some minor additions.
Change-Id: Id577db449ae0e5baec40621d4a08387dbd755342
(a) Don't check that the audiotrack is initialized (will
not be true in static mode until the first write call)
(b) Block until it completes playing.
Change-Id: I50eb6cece00f3d12f3b75102d62014c360ac0346
This adds two methods:
SynthesisRequest.getMaxBufferSize()
SynthesisRequest.completeAudioAvailable()
Change-Id: I1186eed45997ee9a7e51212c8d6706dd324ca949
This removes the old non-public C++ API for TTS
engines and replaces it with a Java API.
The new API is still @hidden, until it has been approved.
Bug: 4148636
Change-Id: I7614ff788e11f897e87052f684f1b4938d539fb7
This is for passing the http referer of a request, e.g., for speech
input from within a browser.
Bug: 4161306
Change-Id: I56fdb7466edd985aab6df8364be1f1619a11a00d
Pass speech recognizer confidence values in the SpeechRecognizer and
RecognizerIntent APIs through to the caller. This change defines new
API constants for keys to these values.
The corresponding change is being made to Google Voice Search, and should be
made to any other implementations.
Bug: 3267513
Bug: 4163206
Change-Id: I294553f2eb9eb3be21298b8434117c8c5309558d
This is just a code clean-up change. It should have no effects
on behavior.
- The exception clean-up code was copied in lots of places,
added a restart() method for it.
- Use Log.e to print exception stack trace instead of doing
it manually.
- Moved return statements outside finally-blocks. Having them
in the finally-block makes no difference, but can be misleading.
Also, Eclipse complains about it.
Change-Id: Ibfb18ca89d81d1fd8d212f29b6f8ab305b52ec49
The NB_CACHED_PARAMS is the number of cached parameters, not
the size of the array for cached parameters.
Change-Id: I3d3053bbf247f3f89ae1ea07a931d33df7857a7c
Bug #3458256
If an error occurs and simultaneously user cancels the recognition, here is what happens:
1. dispatchCancel() is called since the user requested cancel. It passes the first "if" successfully.
private void dispatchCancel(IRecognitionListener listener) {
if (mCurrentCallback == null) {
if (DBG) Log.d(TAG, "cancel called with no preceding startListening - ignoring");
} else if (mCurrentCallback.mListener.asBinder() != listener.asBinder()) {
Log.w(TAG, "cancel called by client who did not call startListening - ignoring");
} else { // the correct state
RecognitionService.this.onCancel(mCurrentCallback);
mCurrentCallback = null;
if (DBG) Log.d(TAG, "canceling - setting mCurrentCallback to null");
}
}
2. Error occurs in the app, which sets the mCurrentCallback to null:
public void error(int error) throws RemoteException {
mCurrentCallback = null;
mListener.onError(error);
}
3. the second "if" is reached in dispatchCancel()
4. boom
Change-Id: I54cdcc98b495d820a2caead1709d8dee968c461e
The creation of a TextToSpeech object causes the TTS service to
start, and the initialization listener to be called when the
service is connected. But the listener is never called when
the service binding failed to be notified of this error.
The fix consists in checking the result of the bind to service
operation, and notify the listener in case of an error.
More log was added in case speak() and synthesizeToFile() are
called but the service is not known to have started.
Change-Id: I7dcc1fa44be31fee3177ec6215fca3306377b934
This CL unhides two "keys" an application can use to specify for
an utterance its playback volume and pan. Those two new keys
use the same mechanism already in place for an application to
specify stream type and utterance ID.
Change-Id: I020363487d7a0f471fb8d3ed739c561d2ab4b0a9
Add two new parameters that are used when synthesizing text and
playing it back directly to control the volume and left-right
panning of the output.
Panning is applied using a balance law, which is not energy-preserving
but which doesn't lower the volume when not panning / panning to
center (legacy behavior).
Reduced amount of logs, and removed spoken text.
In TextToSpeech.java: added convenience method to handle the setting
of the cached synthesis parameters.
Change-Id: I235d3d3193283ccc1891e2065d43787e3f63304d
* commit '0655b31d758ce8fe054c98c6d288e424f2d273e2':
Unhide RecognizerResultsIntent. This API was reviewed for Froyo, but we didn't want to make it public then because it wasn't until our first Market release of Voice Search that the APIs would be in use by our app.
