Unduck AVAudioSession with SKTTSPlayer - ios

Question

It has been a while since I tried to properly mix the SKTTSPlayer with the AVAudioSession and I have many problems that I can't resolve.
I want the SKTTSPlayer to duck the other audio when it is playing and unduck the other audio when it has finished playing.
Normally, with the AVSpeechSynthesizer, it is pretty easy to duck and unduck the other audio, here is a block of code that does this work pretty easily :
override func viewDidLoad() {
super.viewDidLoad()
var audioSession = AVAudioSession.sharedInstance()
do {
try audioSession.setCategory(AVAudioSessionCategoryPlayback, withOptions: .DuckOthers)
} catch {
// TODO: Handle error
}
synth.delegate = self
speak("I really want to unduck this session")
}
func setAudioSessionActive(beActive: Bool) {
do {
try AVAudioSession.sharedInstance().setActive(beActive)
print("Setting AVAudiosession active: ", beActive)
} catch let error as NSError {
print("Setting AVAudiosession state failed: \(error.description)")
}
}
func speak(string: String) {
let utterance = AVSpeechUtterance(string: string)
utterance.rate = AVSpeechUtteranceDefaultSpeechRate / 2
utterance.voice = AVSpeechSynthesisVoice(language: language)
synth.speakUtterance(utterance)
}
func speechSynthesizer(synthesizer: AVSpeechSynthesizer, didFinishSpeechUtterance utterance: AVSpeechUtterance) {
if (synthesizer.speaking) {
return
}
setAudioSessionActive(false)
}
As you can see, this can do the work pretty easily, you activate the session once and then everytime an utterance is finished, you desactivate the seesion.
This is not so simple with the SKTTSPlayer. Even if it seems that the SKTTSPlayer is an object of AVSpeechSynthesizer modified by Skobbler, you cannot unduck a session by just desactivate it. If you does that, the session will unDuck and after Duck again.
Here is a block of code that can reproduce the error :
override func viewDidLoad() {
super.viewDidLoad()
var audioSession = AVAudioSession.sharedInstance()
do {
try audioSession.setCategory(AVAudioSessionCategoryPlayback, withOptions: .DuckOthers)
} catch {
// TODO: Handle error
}
let advisorSettings = SKAdvisorSettings()
advisorSettings.language = SKAdvisorLanguage.EN_US
advisorSettings.advisorType = SKAdvisorType.TextToSpeech
SKRoutingService.sharedInstance().advisorConfigurationSettings = advisorSettings
let ttsSettings = SKAdvisorTTSSettings()
ttsSettings.rate = 0.08
ttsSettings.pitchMultiplier = 0.8
ttsSettings.volume = 1
ttsSettings.preUtteranceDelay = 0.1
ttsSettings.postUtteranceDelay = 0.1
SKTTSPlayer.sharedInstance().textToSpeechConfig = ttsSettings
SKTTSPlayer.sharedInstance().delegate = self
speak("I really want to unduck this session")
}
func setAudioSessionActive(beActive: Bool) {
do {
try AVAudioSession.sharedInstance().setActive(beActive)
print("Setting AVAudiosession active: ", beActive)
} catch let error as NSError {
print("Setting AVAudiosession state failed: \(error.description)")
}
}
func speak(string: String) {
SKTTSPlayer.sharedInstance().playString(string, forLanguage: SKAdvisorLanguage.EN_US)
}
func speechSynthesizer(synthesizer: AVSpeechSynthesizer, didFinishSpeechUtterance utterance: AVSpeechUtterance) {
setAudioSessionActive(false)
}
func TTSPlayer(TTSPlayer: SKTTSPlayer!, willPlayUtterance utterance: AVSpeechUtterance!) {
NSLog("will play")
}
The only workaround I found for this problem is to create a blanksound with an AVAudioPlayer. This sound is played everytime an utterance is finished and the session is unDuck when the blanksound is finished.
override func viewDidLoad() {
super.viewDidLoad()
var audioSession = AVAudioSession.sharedInstance()
do {
try audioSession.setCategory(AVAudioSessionCategoryPlayback, withOptions: .DuckOthers)
} catch {
// TODO: Handle error
}
let advisorSettings = SKAdvisorSettings()
advisorSettings.language = SKAdvisorLanguage.EN_US
advisorSettings.advisorType = SKAdvisorType.TextToSpeech
SKRoutingService.sharedInstance().advisorConfigurationSettings = advisorSettings
let ttsSettings = SKAdvisorTTSSettings()
ttsSettings.rate = 0.08
ttsSettings.pitchMultiplier = 0.8
ttsSettings.volume = 1
ttsSettings.preUtteranceDelay = 0.1
ttsSettings.postUtteranceDelay = 0.1
SKTTSPlayer.sharedInstance().textToSpeechConfig = ttsSettings
SKTTSPlayer.sharedInstance().delegate = self
speak("I really want to unduck this session")
}
func setAudioSessionActive(beActive: Bool) {
do {
try AVAudioSession.sharedInstance().setActive(beActive)
print("Setting AVAudiosession active: ", beActive)
} catch let error as NSError {
print("Setting AVAudiosession state failed: \(error.description)")
}
}
func speak(string: String) {
SKTTSPlayer.sharedInstance().playString(string, forLanguage: SKAdvisorLanguage.EN_US)
}
func speechSynthesizer(synthesizer: AVSpeechSynthesizer, didFinishSpeechUtterance utterance: AVSpeechUtterance) {
initializeAudioPlayer()
}
func TTSPlayer(TTSPlayer: SKTTSPlayer!, willPlayUtterance utterance: AVSpeechUtterance!) {
NSLog("will play")
}
func initializeAudioPlayer(){
let blankSoundURL = NSBundle.mainBundle().pathForResource("blankSound", ofType: "mp3")
if !NSFileManager.defaultManager().fileExistsAtPath(blankSoundURL!)
{
return
}
else
{
do {
audioPlayer = try? AVAudioPlayer(contentsOfURL: NSURL(fileURLWithPath: blankSoundURL!), fileTypeHint: nil)
audioPlayer.delegate = self
audioPlayer.prepareToPlay()
audioPlayer.play()
}
}
}
func audioPlayerDidFinishPlaying(player: AVAudioPlayer, successfully flag: Bool) {
setAudioSessionActive(false)
}
This is, I think, pretty difficult for nothing and it makes the AVAudioSession hard to work with.
Do you guys have any ideas why I can't unDuck a session with the SKTTSPlayer in the normal way ?
Thank you for reading, I know this is a pretty long post, but I think it might help someone.

Answer 1

Using the TTS option
Set our audio engine to TTS (this way you'll receive the text that needs
to be played)
Implement the didUpdateFilteredAudioInstruction callback ­ here is where
the text instruction is received.
At this time you will have the play the instruction using your custom
TTS engine ­ this is implementation specific to your 3'rd party engine
Return false (this will stop the default tts engine from playing the
instruction again)
The end code will look something like this:
(BOOL)routingService:(SKRoutingService )routingService
didUpdateFilteredAudioInstruction:(NSString)instruction
forLanguage:(SKAdvisorLanguage)language {
//your code goes here - you need to play the "instruction" in a
certain "language"
return false;
}
Using your own AVSpeechSynthesizer:
If it's a TTS engine issue then the only way of addressing this would be
to change the TTS engine you are using (we have limited control over it ­
see
http://developer.skobbler.com/docs/ios/2.5.1/Classes/SKAdvisorTTSSettings.h
tml ). See also the iOS API documentation:
https://developer.apple.com/library/prerelease/ios/documentation/AVFoundati
on/Reference/AVSpeechSynthesizer_Ref/index.html