Telnyx Ultra Voices

Use Telnyx Ultra for emotionally expressive, multilingual text-to-speech with sub-100ms time to first byte.

Telnyx Ultra is a premium text-to-speech model that delivers natural, emotionally expressive speech across 36 languages. With sub-100ms time to first byte and support for SSML emotion tags, Ultra is built for real-time AI assistants and customer-facing voice experiences where quality and responsiveness matter.

What makes Ultra different

Feature	Natural / NaturalHD	Ultra
Time to first byte	~200–300ms	< 100ms
Expressive mode	—	Automatic emotional interpretation with SSML emotion tags
Nonverbal cues	—	[laughter] and other natural vocalizations
Languages	Varies by voice	36 languages

Voice format

Ultra voices use the format:

Telnyx.Ultra.<voice_id>

Browse available Ultra voices on the Available Voices page and filter by the Ultra model.

Expressive mode for AI Assistants

When using Ultra voices with AI Assistants, you can enable expressive mode — a toggle that lets the AI model dynamically control the emotional delivery of speech during a live conversation.

How it works

With expressive mode enabled, the assistant’s system prompt is automatically augmented with instructions for using SSML emotion tags. The AI model then decides — in real time — when and how to apply emotional expression based on the conversation context.

For example, the assistant might:

• Use an excited tone when delivering good news.
• Respond with empathy when a customer describes a problem.
• Insert [laughter] for natural, friendly moments.
• Speak neutrally for matter-of-fact information.

This “director-style” approach means you don’t need to hard-code emotions into your prompt — the AI interprets the emotional subtext naturally, with optional explicit guidance for important moments.

Demo

Here’s an example conversation with an Ultra voice assistant using expressive mode. Notice how the assistant automatically selects the appropriate emotion — [angry], [sad], [happy] — and uses [laughter] naturally, all without any explicit instructions in the prompt:

Enable in the portal

1. Go to your assistant in the Telnyx Portal.
2. Under Voice Settings, select a Telnyx Ultra voice.
3. Toggle Expressive Mode on.
4. Save your assistant.

Enable via API

Set expressive_mode: true in your assistant’s voice_settings:

curl -X PATCH "https://api.telnyx.com/v2/ai/assistants/YOUR_ASSISTANT_ID" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "voice_settings": {
      "voice": "Telnyx.Ultra.YOUR_VOICE_ID",
      "expressive_mode": true
    }
  }'

expressive_mode is only supported for Telnyx Ultra voices (voices starting with Telnyx.Ultra.). Enabling it with a non-Ultra voice will return a validation error.

SSML emotion reference

When expressive mode is enabled, the assistant can use these SSML emotion tags in its responses. You can also use these tags with Ultra voices in Call Control speak commands.

Format

Place an <emotion> tag before the text you want to affect:

<emotion value="EMOTION" />Your text here.

Available emotions

Primary emotions (best quality):

Tag	Use for
<emotion value="angry" />	Firm, assertive delivery
<emotion value="excited" />	Energetic, enthusiastic responses
<emotion value="content" />	Calm, satisfied tone
<emotion value="sad" />	Empathetic, subdued delivery
<emotion value="scared" />	Anxious, worried tone

Additional emotions:

happy, enthusiastic, curious, calm, grateful, affectionate, sarcastic, surprised, confident, hesitant, apologetic, determined, frustrated, disappointed, and more.

Nonverbalisms

Insert [laughter] directly in the text for natural laughing:

That's a great joke! [laughter] Okay, let me help you with that.

You don’t need emotion tags for every sentence. Ultra naturally interprets emotional subtext from the text itself — use explicit tags sparingly for ambiguous readings or important moments, like a director guiding an actor. Omitting the tag is equivalent to neutral delivery and produces the most natural speech.

Language support

Ultra supports 36 languages via the language_boost setting, which improves pronunciation accuracy for the target language:

Arabic, Bengali, Bulgarian, Chinese, Czech, Danish, Dutch, English, Finnish, French, German, Gujarati, Hebrew, Hindi, Indonesian, Italian, Japanese, Korean, Malay, Marathi, Māori, Norwegian, Polish, Portuguese, Punjabi, Romanian, Russian, Slovak, Spanish, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, and Vietnamese.

Set language_boost in your assistant’s voice settings to improve pronunciation for the target language.

Next steps

Available Voices

Browse Ultra voices and filter by language and gender.

Voice Design Lab

Design custom voices from natural language prompts.

AI Assistants

Build voice AI assistants using Ultra with expressive mode.

Call Control TTS

Use Ultra voices with Call Control speak commands.

TTS WebSocket Streaming

Stream text to the TTS API and receive audio in real time.