RT-Voice - Run-time text-to-speech solution

Hi Stefan,

I facing a similar issue when I try to add SSML to my text from script.
This is the code is I am writing in script:
voiceOutput = "<s>I am now < prosody rate = \"x -slow\" > speaking at half speed.</ prosody >"+
"I am now<prosody rate= \"slow\" > speaking at 2 / 3 speed.</ prosody >"+
"I am now<prosody rate= \"medium\" > speaking at normal speed.</ prosody >"+
"I am now<prosody rate= \"fast\" > speaking 33 % faster.</ prosody ></s>";
Speaker.Speak(voiceOutput, null, Speaker.VoiceForName(Voice));

It works fine when I use it directly in the RTV text box.
Also works fine for break strength property in SSML through script.

But in other cases as the above example the sentences are read correctly but the SSML tags are ignored.
How to solve it?
Thanks in advance.

 

Hi

I assume your builds are for Windows?
Because I just did a test and after some minor tweaks to your text examples, it works.
Here is the code:

Code (CSharp):

1. using Crosstales.RTVoice;
2. using UnityEngine;
3.  
4. public class RTVTest : MonoBehaviour
5. {
6.     public string Voice = "Zira";
7.  
8.     [TextArea]
9.     public string VoiceOutput = "<s>I am now <prosody rate=\"x-slow\"> speaking at half speed.</prosody>"+
10.                          "I am now<prosody rate=\"slow\"> speaking at 2 / 3 speed.</prosody>"+
11.                          "I am now<prosody rate=\"medium\"> speaking at normal speed.</prosody>"+
12.                          "I am now<prosody rate=\"fast\"> speaking 33 % faster.</prosody></s>";
13.    
14.     public void OnEnable()
15.     {
16.         Speaker.OnVoicesReady += onVoicesReady;
17.     }
18.  
19.     public void OnDisable()
20.     {
21.         Speaker.OnVoicesReady -= onVoicesReady;
22.     }
23.  
24.     public void Speak()
25.     {
26.         Speaker.Speak(VoiceOutput, null, Speaker.VoiceForName(Voice));
27.     }
28.  
29.     private void onVoicesReady()
30.     {
31.         Speak();
32.     }
33. }

Btw, it's important to wait for the "OnVoicesReady"-callback before calling RT-Voice.

I hope this helps you further, please let me know.


Cheers
Stefan

 

Thanks for the quick reply.

Yes, the builds are for Windows.

According to my understanding OnVoicesReady is to check if the voices in unity have been initialized and set and if we try playing the speech text before that, then there wont be any voice heard.

But in my case, I was able to hear the voice, just the SSML tags were ignored.

Anyways found the issue in my case.
It was due to additional empty spaces in my tags in the script. (silly from my part )
Thanks for the help

 

I would like to ask if you have a suggestion for a use case. For now i have an avatar with SALSA and RT Voice. I also added a Text to Speech plug in from SALSA in order to provide a text for the avatar to speak. What i would like to do is to use RT Voices with Google Text To Speech plug in. (and of course to have the avatar with salsa working with the same way as before) Is that possible and if yes can you suggest general instructions? Not with details (for now) Just a basic technical thought

 

Hi

You just have to install the following plugin:
https://assetstore.unity.com/packages/slug/115170?aid=1011lNGT

Afterwards, install our integration package from the "3rd party"-folder inside the asset.
Then you have to configure the plugin with you Google Cloud API-key and your good to go.

Does that help you further?


Cheers
Stefan

 

Yes it was very helpfull! Thank you again Stefan.

 

I'm sure this is something dumb but I having an issue getting my string into speechText.cs . After I set the string, it will throw the following error unless click in the text field to force OnValidate :

Could not start process: System.ComponentModel.Win32Exception (0x80004005): ApplicationName='say', CommandLine=' -v "Samantha" -r 169 -o "/var/folders/wj/tcqktw9s2l921j8khcs4fwy80000gp/T/DefaultCompany/Reader/rtvoice_a8b66ecc-9cf8-4952-958a-c22331661368.aiff" --file-format=AIFFLE " ////// 'my big "string" here.' //////"

I'm doing this:

Code (CSharp):

1. speechText.Text = myParsedParagraph;
2. speechText.Speak();

My suspicion is it has something to do with \" characters in my string but I'm not sure.

 

Is your text " ////// 'my big "string" here.' //////"?

What's your Unity and RT-Voice version?

 

no it's a long string (around 1700 characters) that was joined from multi lines.

 

Is it possible that you send me the text for further analysis?

 

Wow thanks, I have emailed...

 

My game is for Android and iOS. So, these questions pertain to those platforms.

Pre-purchase Questions:

• Does it work with AAB and IL2CPP?
• Does the user have to manually change any settings on their device, or will it just work?
• What languages/cultures are supported? Is there a list somewhere that I could take a look at? Is it different for each device?
• In the documentation there is a method for errors. Do you have a list of potential errors? What is the most common error?

Thanks

 

Hi

Thank you for your interst in RT-Voice!

Here are the answers to your questions:

1. Yes, RT-Voice works with IL2CPP and AAB
2. It just works with the installed voices; no changes needed
3. iOS has the same voices on every device; Android is mostly Google TTS but can slightly differ due the fragmented versions
4. Errors are very rare and will most likely never occur on mobile

I hope this helps you further,

Cheers
Stefan

 

https://www.crosstales.com/media/data/assets/rtvoice/RTVoice-doc.pdf


I added the VoiceInitalizer to the first scene. I clicked to initialize all voices. As soon as the scene loads, it keeps saying "Crosstales" with each voice.

• Is there a way to initialize the voices without saying "Crosstales"?
• Am I supposed to add the initializer to each scene or only the first scene?

Code (CSharp):

1. Speaker.SpeakNative("Hello World!", Speaker.VoiceForCulture("en"));

I would like to use a female gender by default with SpeakNative. How would I do that?

 

Hi

"VoiceInitalizer" must only be used when you encounter any problems with voices, like long response times.
If you really need it, it should be added to the first scene.

To use a specific gender, do it like that:

Code (CSharp):

1. Speaker.SpeakNative("Hello World!", Speaker.VoiceForGender(Model.Enum.Gender.FEMALE, "en"));

I hope this helps you further.


Cheers
Stefan

 

I copied and pasted the code but it says "The name 'Model' does not exist in the current context."

 

Can you please try "Crosstales.RTVoice.Model.Enum.Gender.FEMALE"?

 

It works.

 

I put the RTVoice prefab on my Map scene, but it gets deleted for no apparent reason.

Everything is working properly:


The RTVoice prefab disappeared for no reason:


I figured out how to reproduce the error:

• Add RTVoice prefab to Map scene.
• Change to a different scene.
• Go back to the Map scene and the RTVoice prefab is gone.

Do you know why this is happening and how I can fix it?

 

The last picture shows it's not gone; RT-Voice is a singleton and has "DontDestroyOnLoad" enabled. When you hit play, it seems to be gone, but it's under "DontDestroyOnLoad" in the scene.

Does that answer your question?

 

I uploaded that last picture by mistake. Sorry for the confusion.

I will figure out how to record my screne so that I can show you exactly what is happening. Thanks for the quick support. I appreciate it!

 

Hmm, that sounds strange. Did you hit save before switching the scene?


Edit:
Adding the prefab via menu won't set the scene dirty and it won't be marked as "changed".
Therefore, Unity will not ask to save the changes and you have to do it manually.
However, this will be fixed in the next version.

 

edit

 

Yes, I tried saving the scene and also the project.

 

Please add the prefab manually from "Assets/Plugins/crosstales/RTVoice/Prefabs" and try it again.

 

I did that and it is working now. Do you know what was happening in case I run into this problem again? I have a feeling that this is a bug in Unity, but what do you think?

Thanks for the quick support. I appreciate it!

 

It will be fixed with the next release; just remember until then

 

Hi Crosstales,

We've been using the old RTVoice for our project for a while and today decided to update/upgrade.
When Importing, RTVoice seemed to have caused a bunch of errors (they all seem the same but for different scripts). Here's one of them:
Assets/Plugins/crosstales/Common/Editor/Util/BaseEditorHelper.cs(402,22): error CS0121: The call is ambiguous between the following methods or properties: `Crosstales.ExtensionMethods.CTEquals(this string, string, System.StringComparison)' and `Crosstales.ExtensionMethods.CTEquals(this string, string, System.StringComparison)'

Could you please help with resolving this error?

Thanks!

 

I don't know if this helps or not:

https://www.crosstales.com/media/data/assets/rtvoice/RTVoice-doc.pdf

 

Hi

Yes, the chapter 10 of the documentation is normally the way.
But your version of our asset seems quite old, therefore, I recommend creating an empty scene, delete the whole "Assets/Plugins/crosstales"-folder and import the asset from scratch.


Cheers
Stefan

 

Code (CSharp):

1. Speaker.SpeakNative("Hello World!", Speaker.VoiceForGender(Model.Enum.Gender.FEMALE, "en-GB", 0, "en"));

The above code causes the voice to fallback to Alice (it-IT, FEMALE). This is an Italian voice.



My device has "en-US" and "en-AU" in a female voice.

How can I ensure that the fallback is both female and English?

 

Are you using macOS or iOS as platform?
Which is the version of the asset and Unity?

 

My game will be developed for Android and iOS. I am testing through the Unity editor on a Mac Mini.

Asset Version: 29.01.2020
Unity Version: 2019.2.18

 

Do you wait for the OnVoicesReady-callback before executing the "Speak"-method?

Here is an example:

Code (CSharp):

1. using UnityEngine;
2. using Crosstales.RTVoice;
3.  
4. public class SpeakWhenReady : MonoBehaviour {
5.  
6.     public void OnEnable()
7.     {
8.         // Subscribe event listeners
9.         Speaker.OnVoicesReady += onVoicesReady;
10.     }
11.  
12.     public void OnDisable()
13.     {
14.         // Unsubscribe event listeners
15.         Speaker.OnVoicesReady -= onVoicesReady;
16.     }
17.  
18.     private void onVoicesReady()
19.     {
20.         Speaker.SpeakNative("Hello World!", Speaker.VoiceForGender(Model.Enum.Gender.FEMALE, "en-GB", 0, "en"));
21.     }
22. }
23.  

 

I can hear that the voice is Italian.

 

I just tested it on macOS Catalina and my voice is "Fiona", a Scottish voice.
So it works on my side; can you please get the voice in your "TTS"-method as variable and print it?

 

I don't know what changed, but I am now getting Fiona as the fallback voice.

 

The fallback is working properly with "en-GB":


The fallback doesn't work with "en-CA". I am getting the Italian voice again. There is no alert message in the console.

Can you try "en-CA" to see if the fallback works for you?

PS. I just tested it on Android. I am getting the wrong voice for "en-CA". It sounds like a French voice. The fallback isn't working.

 

VoiceForGender has a fifth parameter, called "isFuzzy", which is optional and set to true by default.
Setting it to false will solve the issues you encounter:

Code (CSharp):

1. Model.Voice v = Speaker.VoiceForGender(Model.Enum.Gender.FEMALE, "en-GB", 0, "en", false);

We will change the parameter to false as default in the next release.

 

I'm using the demo SpeechText components along with the Audio Source component for reverb. All works great in the Editor, but in my iOS build there are two issues:
1. I don't hear the reverb fx
2. calling the method TextSpeech.Silence() does not appear to work, though it works in the editor
Any idea what could be causing this?

 

Hi Jason

The TTS from iOS doesn't support audio file generation and therefore, no Unity audio effects are possible.
If you need them on iOS, please use another (online) provider like MaryTTS, AWS Polly etc.

About the issue on iOS: what's the iOS version on your device?


Cheers
Stefan

 

Ah ok. I'm testing on iOS 12.4.1 / RTvoice 2020.1.1 / Unity 2018.4.14

 

Please send us your invoice via email and we will give you access to the latest version.

 

Hi Stefan,

We are using RT-Voice and we need to play different voices depending on the customer's language. In my PC there are four Microsoft voices installed (Spanish female, English male and female, Portuguese female) but every time it is played the same: Spanish female.

RT-Voice version: 2.9.8
Unity version: 2018.4.14f
OS: Windows 10

I have been debugging and it seems that RT-Voice is detecting the four voices right (Speaker.Voices has the right values), but calling Speaker.Speak with the desired voice is not giving me the expected results. Looking deeper into the code, in VoiceProviderWindow the args string passed to the speakToFileProcessprocess thread call is the following (the dir argument has been shortened for clarity and privacy):

--speakToFile "<prosody pitch='+0%'>Esta é uma frase de exemplo, para ouvir no simulador de direção.</prosody>" "C:/...104cb3b9-e20d-4fbc-9634-e7f880701ee6.wav" 0 100 "Microsoft Maria Desktop"

But even if the last parameter has the right value (Maria is the Portuguese voice) , I am hearing the Spanish voice.

I have also tried to change the default voice in Windows but the result is the same. Do you have any advice? Thanks in advance,

Aarón.

 

Hi Aaron

That's because not every voice has SSML support. Enable "Auto Clear Tags" on the "Speaker"-component to select the right voice.
We also strongly recommend updating to the latest version of RT-Voice PRO, since 2.9.8 and the DLL version are no longer supported.


Cheers
Stefan

 

Hello,

Thank you for your quick answer. Sadly enabling "Auto Clear Tags" makes no difference in this case. Regardless of which voice I choose, I hear "You have selected the [right_voice_name]" with the same not-selected voice in the Test.

Since enabling "Auto Clear Tags" should solve any SSML compatibility issues, sould I discard SSML compatibility as the problem? Any tip would be really appreciated. Thanks again!

 

I really can't say, it's been a long time since the version you are using.
It will work with the current version, so please consider updating.

 

Ok, I will update and let you know.

 

As you said, after updating it worked. Thanks a lot.

 

The first time that the audio is played, it reads properly. The second time, it spells the word. For example, it would spell "v", "o", "i", "c", "e", instead of saying "voice".

Edit: I think I figured it out. The string had a misspelled word in it.