Stream Speech
Synthesize speech with ultra-low latency over a streaming connection.
Choose the Base URL from the URL dropdown (Global URL or a pinned Region)
Note: Global URL auto-routes to the nearest region.
Headers
Request
The text to speak, for example Hello there, friend.
To control how a word is pronounced, replace it with its IPA, wrapped in forward slashes and double quotes:
Welcome to "/mɜːrf/".
In a raw JSON body, escape the quotes: \"/mɜːrf/\". To apply the same pronunciation every time a word appears, use pronunciationDictionary instead. Learn more in Pronunciations.
Use the GET /v1/speech/voices API to find supported voiceIds. You can use either the voiceId (e.g. en-US-natalie) or just the voice actor’s name (e.g. natalie).
The model to use for audio output. Valid values: falcon-2.
Valid values: STEREO, MONO
Format of the generated audio file.Valid values: MP3, FLAC, WAV, ALAW, ULAW, OGG, PCM
Specifies the language for the generated audio, enabling a voice to speak in multiple languages natively. Valid values: “en-US”, “en-UK”, “es-ES”, etc. Use the GET /v1/speech/voices endpoint to retrieve the list of available voices and languages.
Custom pronunciations for words in text. Each key is a word as it appears in the text. Each value has a type and a pronunciation:
IPA: the IPA written without quotes or slashes, for examplelaɪv.SAY_AS: the words to say instead, for exampletwo thousand and ten.
An entry applies every time its word appears in the request.
Example:
{
"live": { "type": "IPA", "pronunciation": "laɪv" },
"2010": { "type": "SAY_AS", "pronunciation": "two thousand and ten" }
}
Adjusts loudness relative to the model’s default output, from -50 to 50. Positive values increase level; negative values decrease it; 0 leaves audio as generated. Values outside the range are clamped to the nearest limit.
This field is superseded by locale field. Please migrate to locale field to ensure compatibility with future API versions.
Specifies the language for the generated audio, enabling a voice to speak in multiple languages natively. Valid values: “en-US”, “en-UK”, “es-ES”, etc. Use the GET /v1/speech/voices endpoint to retrieve the list of available voices and languages.