How to change speech speed(rate) when use "Input text streaming" in text to speech?

Question

How to change speech speed(rate) when use "Input text streaming" in text to speech?

datou ai 0

I have tried to set property(SpeechSynthesisReqest_Rate) but it doesn't work.

speech_config = speech_sdk.SpeechConfig(
    endpoint=endpoint,
    subscription=AZURE_TTS_API_KEY,
)
speech_config.set_property(
    speech_sdk.PropertyId.SpeechSynthesisRequest_Rate, "1.2"
 )
speech_config.set_speech_synthesis_output_format(
    speech_sdk.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus
)
speech_synthesizer = speech_sdk.SpeechSynthesizer(
    speech_config=speech_config, audio_config=None
)
speech_synthesizer.synthesizing.connect(self.speech_event_callback)
tts_request = speech_sdk.SpeechSynthesisRequest( 
  input_type=speech_sdk.SpeechSynthesisRequestInputType.TextStream
)

datou ai 0 Reputation points

2026-01-19T04:34:50.2566667+00:00

So does SSML streaming mean sending multiple complete SSML segments in a stream? For example, sending:

<prosody rate="1.2">hello </prosody>

<prosody rate="1.2">world.</prosody>

<prosody rate="1.2">Today is </prosody>

<prosody rate="1.2">a nice day!</prosody>

datou ai 0

I couldn’t find an SsmlStream enum. In the SDK, SpeechSynthesisRequestInputType only includes the following values:

class SpeechSynthesisRequestInputType(Enum):
    """
    Defines the input type of speech synthesis request
    """

    NoneInput = 0

    Text = 1
    """
    Plain text input
    """

    SSML = 2
    """
    SSML (Speech Synthesis Markup Language) input
    """

    TextStream = 3
    """
    Text stream input
    """

Chakaravarthi Rangarajan Bhargavi 1,205 Reputation points MVP

2026-01-19T06:12:33.5666667+00:00
There is no SsmlStream input type in the Azure Speech SDK. This is expected behavior. The SDK only exposes the following input types: text to one plain text, ssml to one-time SSML, textstream to streaming input (plain text or SSML).

What’s often referred to as “SSML streaming” is not a separate enum or feature, but a pattern implemented using TextStream.

When you use SpeechSynthesisRequestInputType.TextStream, you can stream multiple SSML chunks, but each chunk must be a complete, valid SSML document wrapped in <speak>...</speak>.

So instead of streaming partial SSML tags, you send full SSML segments one after another.

Each write to the stream is processed independently by the service.

SsmlStream does not exist (by design). Use TextStream to stream complete SSML documents, Control speech behavior only through SSML. This is the officially supported and expected approach

Example:

<speak><prosody rate="1.2">Hello</prosody></speak> <speak><prosody rate="1.2">world.</prosody></speak> <speak><prosody rate="1.2">Today is a nice day.</prosody></speak>
Chakaravarthi Rangarajan Bhargavi 1,205 Reputation points MVP

2026-01-20T02:58:54.5033333+00:00

Hi, Did you had a chance to check on the same and is it working? if its working shall i convert it to solution?

1 answer

Your answer

datou ai 0 Reputation points

2026-01-19T04:34:50.2566667+00:00

So does SSML streaming mean sending multiple complete SSML segments in a stream? For example, sending:

<prosody rate="1.2">hello </prosody>

<prosody rate="1.2">world.</prosody>

<prosody rate="1.2">Today is </prosody>

<prosody rate="1.2">a nice day!</prosody>
datou ai 0 Reputation points

2026-01-19T04:38:34.66+00:00

I couldn’t find an SsmlStream enum. In the SDK, SpeechSynthesisRequestInputType only includes the following values:

class SpeechSynthesisRequestInputType(Enum): """ Defines the input type of speech synthesis request """ NoneInput = 0 Text = 1 """ Plain text input """ SSML = 2 """ SSML (Speech Synthesis Markup Language) input """ TextStream = 3 """ Text stream input """
Chakaravarthi Rangarajan Bhargavi 1,205 Reputation points MVP

2026-01-19T06:12:33.5666667+00:00

There is no SsmlStream input type in the Azure Speech SDK. This is expected behavior. The SDK only exposes the following input types: text to one plain text, ssml to one-time SSML, textstream to streaming input (plain text or SSML).

What’s often referred to as “SSML streaming” is not a separate enum or feature, but a pattern implemented using TextStream.

When you use SpeechSynthesisRequestInputType.TextStream, you can stream multiple SSML chunks, but each chunk must be a complete, valid SSML document wrapped in <speak>...</speak>.

So instead of streaming partial SSML tags, you send full SSML segments one after another.

Each write to the stream is processed independently by the service.

SsmlStream does not exist (by design). Use TextStream to stream complete SSML documents, Control speech behavior only through SSML. This is the officially supported and expected approach

Example:

<speak><prosody rate="1.2">Hello</prosody></speak> <speak><prosody rate="1.2">world.</prosody></speak> <speak><prosody rate="1.2">Today is a nice day.</prosody></speak>
Chakaravarthi Rangarajan Bhargavi 1,205 Reputation points MVP

2026-01-20T02:58:54.5033333+00:00

Hi, Did you had a chance to check on the same and is it working? if its working shall i convert it to solution?

Answer 1

Welcome to Microsoft Q&A,

What you’re seeing is actually expected behavior when using input text streaming with Azure Text-to-Speech.

When you use:

SpeechSynthesisRequestInputType.TextStream

the service starts synthesizing audio immediately as text arrives. Because of that, request-level properties like SpeechSynthesisRequest_Rate are ignored. This isn’t a configuration issue in your code, the rate setting simply isn’t supported for text streaming today.

That property works only when the full text is known in advance.

If you need to control speaking speed, the only supported option is to use SSML prosody instead of plain text. In other words, you need to stream SSML, not raw text.

So instead of TextStream, switch to SsmlStream and wrap your text like this:

<prosody rate="1.2">Your text here</prosody>

Once you do that, the speech rate will be applied correctly.

Unfortunately, there’s no way right now to dynamically change speech speed for TextStream inputs, and the SDK doesn’t expose a client-side playback rate control for cloud TTS either. This is a current service limitation rather than a bug.

If this behavior is important for your scenario, SSML streaming is the recommended workaround.

Let us know if you need any further assistance, we’d be happy to help.

Best regards,

Chakravarthi Rangarajan Bhargavi

If you find this answer helpful, please click Accept Answer and upvote. This helps others in the community find the solution more easily.

Share via

How to change speech speed(rate) when use "Input text streaming" in text to speech?

1 answer

Your answer