Feel free to post any issues or questions about Cognitive TTS service here!

[szhaomsft]

we encourage developers to post issues / questions in this forum.

It is monitored regularly

[szhaomsft]

Welcome post like

1. Question about the API
2. Feedbacks about the service
3. Feature request.
4. Sample request.

[phly95]

It seems the output is limited to 10 minutes of audio (at least using the nural option). What if I want to process a long text file, like a required reading or a chapter of a book?

[bodyzatva]

I´m using voice pt-PT-HeliaRUS with language pt_PT for a chatbot in a project for a client.
We are facing issues when the bot speaks email addresses.
When i send this text using ssmlSpeak:

"O seu e-mail é <say-as interpret-as="characters">[email protected]"

The email is not being spelled.

I tried other voices and languages like : pt-BR-HeloisaRUS pt_BR and en-US-JessaRUS en-US.

Only in the voice "en-US-Jessa24kRUS" he spells the name.

Can you tell me why ?

By the way , we have a workaround to force spelling that is separate the email text with spaces:
"O seu e-mail é <say-as interpret-as="characters">a n a r e b e l o @ s a p o . p t < / say-as>"

Is this a problem with pt-PT language? And how can i have better results spelling emails correctly ?

[shoutbomb]

How do I control the pace of the generated speech. I need to slow it down by 10%.
(en-US, JessaNeural)
X-Microsoft-OutputFormat: riff-24khz-16bit-mono-pcm

[hannabonert]

Hello,
I followed the sample from here, and can successfully send 16 kHz audio to the service, and receive a valid response.
How can I use a sampling rate of 44100 hz?

I have tried both of the below, but I get "InitialSilenceTimeout" for every recording that I try.
connection.setRequestProperty("Content-Type", "audio/raw;encoding=signed-integer;bits=16;rate=44100");
connection.setRequestProperty("Content-Type", "audio/wav; codecs=audio/pcm; samplerate=44100");

As a test, I ran the same recording through at 16 kHz, and got a "RecognitionStatus" of "Success". I then resampled it to 44100 kHz, and I got "InitialSilenceTimeout".

I have an example using the speech SDK that works, but now I need to use 44 kHz audio data with the REST API.

Any advice would be greatly appreciated.

Thank you!

[isabirahmed]

Text to Speech does not set the right pitch if two pitches are set in one request.

Sample SSML:

<speak xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="http://www.w3.org/2001/mstts" xmlns:emo="http://www.w3.org/2009/10/emotionml" version="1.0" xml:lang="en-US"><voice name="en-US-SaraNeural">Much against his will Reddy obeyed. <prosody rate="default" pitch="28%" volume="default">“It isn’t the least bit of use,”</prosody> he grumbled, as he trotted towards the Big River. <prosody rate="default" pitch="28%" volume="default">“There won’t be anything there. It is just a waste of time.”</prosody></voice></speak>

I have a sentence with two parts of it set to pitch=28%.
The first part "It isn’t the least bit of use," sounds more like pitch=8% even though its set to 28%
The second part "There won’t be anything there. It is just a waste of time." sounds correct at pitch=28%

Please note this is happening with all the voices and looks like a major bug.
It only happens when you set more than one sentence of the pitch.

Please test this in US East region.
Sample audio file: https://fliki.ai/share/audio/microsoft-pitch-issue-637b4b26dde64016ddbd2a51

[gchiarapa]

I'm using the Rest API, to synthetize text to speech, but I'd like to know to play the response. Any ideas how to convert and play the response?

My request:

    uri = 'https://brazilsouth.tts.speech.microsoft.com/cognitiveservices/v1';
                    method = 'POST';
                    $http({
                        "method": method,
                        "url": uri,
                        "headers": {
                            "Content-type": "application/ssml+xml",
                            "X-Microsoft-OutputFormat": "audio-16khz-64kbitrate-mono-mp3",
                            "Host":"brazilsouth.tts.speech.microsoft.com",
                            "User-Agent": "",
                            "Authorization": "Bearer " + res.data
                        },
                        data: ''

[Chukarslan]

This pertains to commit: d457a6d
GPT Streaming response based on TTS text splitter for each sentence.

Is it possible to share the Python / Node version of the code?

Thanks!