[hitoshi]

Hi, Halle.

Thank you so much.

So far, I did not found the text of the problem.
If I find that, I will report to you immediately.

I will try the method to force a split on long unpunctuated text streams.

I appreciate all your help.
Hitoshi

[hitoshi]

Hi, Halle.

I gave an extreme example. I know that this case is very rare on the real text.
However, I have ever heard the sentence that was interrupted on our app with NeatSpeech.
(Sorry, I do not remember whether it was what kind of sentence.
It was certain that I made the app speak most slowly.)

Our application deals with a variety of text datas.
It is too difficult to determine the point at which to set the punctuations.

If I set a lot of punctuations for a rainy day(=when it was set to speed=-1.0f),
the engine would unnatural speech in the normal speed.
And yet, I am anxious lest the speech is interrupted.

I suggest the following;
Would you like to provide new methods split “sayWithNeatSpeech:withVoice:” ?
e.g.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[FliteController Class]
-(void)makeSayWithNeatSpeech:(NSString *)statement withVoice:(NeatSpeechVoice *)voiceToUse;
…This is the function of only the creation of the speech data. It does not start the speech.

-(BOOL)sayQueue;
…Say the sentence made on “makeSayWithNeatSpeech:withVoice:”.
If the engine have not created yet, it return No.

[OpenEarsEventsObserverDelegate Protocol]
-(void)fliteDidMakeSpeaking:(BOOL)result;
…This is the notification of “The engine created speech data.”.
If there is a problem with the data created, result = No.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Another benefit of this proposal is also available.
Under present conditions, when we called “sayWithNeatSpeech:withVoice:”,
we need to wait just a moment to begin speech.

If we can create the speech data in advance, at the timing when the user taps the play button
we can play immediately speech data.

Your kind consideration of this matter would be sincerely appreciated.
Best regards,
Hitoshi

P.S.
I’m afraid my expressions may be rude or hard to read, because I’m not so good at English.
I will try hard to learn English.

[hitoshi]

Umm.. This isn’t a bug.
Is that so? I’m sorry to hear that…

I think I should not let the users listen to incomplete speech.
In the next version, I’d like you to to consider that “sayWithNeatSpeech:withVoice:” returns the BOOL value.
(When incomplete voice is made, I wish the sound engine to return NO without reconstructing sound.)

As far as I examined, when becoming incomplete sound, “FliteController.audioPlayer” becomes the following condition.
FliteController.audioPlayer.duration = 20.83sec
FliteController.audioPlayer.data.length = 2,000,044byte

This means that there is possibility to become incomplete data when it reconstructs slowly using short text.
For example, it occurs in the following case;
speed=-1.0f
text=”w w w w w w w w w w w w w w w w w w w w w w w w w w ” (Only 52 characters!)

I think that it is difficult to judge with only text length.
It is appropriate that we judge by the result of creating speech data, isn’t it?

Thank you very much for your understanding.
Hitoshi

[hitoshi]

Hi, Halle.

I see.
I’ll investigate the speed by actually reproducing sample sounds.

Thank you for your qucik reply.

Hitoshi

[Author]

Posts