## TTS Streaming Feature

The _TTS Streaming_ feature delivers synthesized speech to your voice AI Agent as it’s generated, rather than waiting for the entire response to finish processing. This feature enables faster, more natural-sounding conversations. Standard TTS systems wait for the entire text to be processed before playing any audio, which introduces a noticeable delay, especially with longer texts. TTS streaming instead processes text in segments and plays each one back as soon as it’s synthesized.

Some speech providers use `streaming` to describe their API’s connection method, for example, a websocket-based API for submitting multiple texts. This concept is different from TTS streaming in Voice Gateway, which refers to audio playout: how quickly synthesized audio starts playing back to the caller.

## Key Benefits

- **Lower response latency.** Audio playback starts as soon as the first words are synthesized, instead of after the full text is processed.
- **More natural conversation flow.** Reduced time-to-first-audio makes voice AI Agent responses feel closer to real-time human conversation, especially with longer text.

## Prerequisites

- A [voice AI Agent](https://docs.cognigy.com/ai/agents/develop/voice) configured.
- A [speech service](https://docs.cognigy.com/voice-gateway/webapp/speech-services) configured in the Voice Gateway Self-Service Portal.

## Restrictions

- The following speech providers don’t support TTS Streaming: AWS, Google, Nuance, Speechmatics, and Deepgram Flux.
- [TTS vendor fallback](https://docs.cognigy.com/voice-gateway/webapp/applications#add-additional-tts-and-stt-vendor) doesn’t work with TTS Streaming. If TTS streaming is activated, the fallback vendor configuration is ignored.
- For ElevenLabs, Voice Gateway uses the [ElevenLabs streaming API](https://elevenlabs.io/docs/api-reference/text-to-speech/stream). The following parameters aren’t supported and can’t be specified in **Extra Options**:
  - `seed`
  - `previous_request_ids`
  - `next_request_ids`
  - `apply_text_normalization`
  - `apply_language_text_normalization`
  - `use_pvc_as_ivc`
  - `pronunciation_dictionary_locators`

## How to Configure TTS Streaming

To activate streaming for your voice AI Agent, follow these steps:

1. Configure Speech Service in Voice Gateway Self-Service Portal.
   1. In the navigation menu of the Voice Gateway Self-Service Portal, select **Speech**.
   2. On the **Speech services** page, select an existing speech service or [create a new one](https://docs.cognigy.com/voice-gateway/webapp/speech-services) by clicking **Add a speech service** in the upper-right corner.
   3. On the **Add Speech service** page, make sure that the **Use for text-to-speech** option is activated.
   4. Activate the **Enable text-to-speech streaming** option.
   5. _(Optional)_ For ElevenLabs, configure the **Extra Options** field to customize the TTS streaming behavior. See the list of unsupported parameters in the [Restrictions](https://docs.cognigy.com/voice-gateway/streaming#restrictions) section.
   6. Click **Save**.

2. Configure the Set Session Config Node in Cognigy.AI.
   1. Go to the Set Session Config Node in your Flow.
   2. In the **Synthesizer (TTS)** section, select your speech service from the **TTS Vendor** list.
   3. Configure **TTS Language** and **TTS Voice** as needed.

## More Information

- [Speech Services](https://docs.cognigy.com/voice-gateway/webapp/speech-services)
- [TTS and STT vendors](https://docs.cognigy.com/voice-gateway/references/tts-and-stt-vendors)

Last modified on July 8, 2026.