Realtime TTS-2 is live. Built for realtime conversation that feels human. Read the Realtime TTS-2 announcement

TextToSpeech

Get voice preview

Inworld Voice API reference: get a short audio preview of a system or workspace voice. Previews are not billed.

GET/tts/v1/voice:preview

Preview a system or workspace voice

Send GET https://api.inworld.ai/tts/v1/voice:preview with query parameters voice_id and model_id, and the header Authorization: Basic <INWORLD_API_KEY>. Use the exact voiceId returned by List voices. System voice IDs such as Ashley work here.

bash
curl --get 'https://api.inworld.ai/tts/v1/voice:preview' \
  --header "Authorization: Basic $INWORLD_API_KEY" \
  --data-urlencode 'voice_id=Ashley' \
  --data-urlencode 'model_id=inworld-tts-2'

The response is a single JSON object:

json
{"audioContent":"<base64-encoded MP3>"}

Base64-decode audioContent to play or save the MP3. Preview text is selected by the service and cannot be customized. Previews are not billed. To speak your own text, use synthesis.

Previews are not billed, but they carry stricter rate and concurrency limits than synthesis. Cache preview audio rather than calling this endpoint once per listener. A voice picker served to many users should replay a stored MP3 instead of requesting it again. See Billing for your plan's limits.

This preview uses the TTS endpoint and a voice_id query parameter. Do not construct /voices/v1/voices/Ashley:preview or treat a system voice ID as a workspace voice-management resource name. Voice metadata and synthesis use different identifier shapes.

Authorizations

Authorizationstringrequired

Your authentication credentials. For Basic authentication, please populate Basic $INWORLD_API_KEY. You can create a key in one command with the Inworld CLI: inworld workspace add-key.

Query Parameters

voice_idstringrequired

The identifier of the voice to preview. Use the List Voices endpoint to discover available voice IDs.

model_idstringrequired

The identifier of the TTS model to use for generating the preview. See Models for available models.

Response

200 - application/json

audioContentstring

The audio data bytes encoded as MP3. The preview text is determined server-side and cannot be customized.