Stream speech (real-time TTS) | EmpirioLabs AI Docs

Streaming variant of /v1/audio/speech. Returns Server-Sent Events as the model synthesizes, with sub-130ms time-to-first-byte on Inworld TTS Mini, sub-250ms on Max. Use this for voice agents and other interactive playback paths.

Event types:

audio.chunk: base64-encoded audio bytes with seq, format, sample_rate, and elapsed_ms fields. Concatenate client-side for progressive playback.
audio.timestamps: emitted when timestamp_type is WORD or CHARACTER, carrying per-token timing info.
audio.done: final event with the assembled hosted URL, total duration, and usage payload for billing.
audio.error: only on upstream errors; followed by [DONE].
[DONE]: terminator (standard SSE).

Currently supported on Inworld TTS Mini / Max. Other TTS models use the synchronous /v1/audio/speech endpoint. Cancelled streams are billed for the characters synthesized up to the cancellation point.

Streaming variant of `/v1/audio/speech`. Returns Server-Sent Events as the model synthesizes, with sub-130ms time-to-first-byte on Inworld TTS Mini, sub-250ms on Max. Use this for voice agents and other interactive playback paths. Event types: * `audio.chunk`: base64-encoded audio bytes with `seq`, `format`, `sample_rate`, and `elapsed_ms` fields. Concatenate client-side for progressive playback. * `audio.timestamps`: emitted when `timestamp_type` is `WORD` or `CHARACTER`, carrying per-token timing info. * `audio.done`: final event with the assembled hosted URL, total duration, and usage payload for billing. * `audio.error`: only on upstream errors; followed by `[DONE]`. * `[DONE]`: terminator (standard SSE). Currently supported on Inworld TTS Mini / Max. Other TTS models use the synchronous `/v1/audio/speech` endpoint. Cancelled streams are billed for the characters synthesized up to the cancellation point.

Authentication

AuthorizationBearer

Pass your EmpirioLabs API key as a bearer token. The Anthropic-style x-api-key header is also accepted on every endpoint.

Request

This endpoint expects an object.

modelstringRequired

inputstringRequired

Text to synthesize. Max 3000 characters per request.

voicestringOptional

voice_idstringOptional

Free-form voice ID; use to address the full 271-voice catalog.

languagestringOptional

output_formatenumOptionalDefaults to WAV

Allowed values:

sample_rateenumOptionalDefaults to 24000

speeddoubleOptionalDefaults to 1

temperaturedoubleOptionalDefaults to 1

timestamp_typeenumOptionalDefaults to NONE

Allowed values:

Response

Server-Sent Events stream (text/event-stream).

Event types:

audio.chunk: base64-encoded audio bytes with seq, format, sample_rate, and elapsed_ms fields. Concatenate client-side for progressive playback.
audio.timestamps: emitted when timestamp_type is WORD or CHARACTER, carrying per-token timing info.
audio.done: final event with the assembled hosted URL, total duration, and usage payload for billing.
audio.error: only on upstream errors; followed by [DONE].
[DONE]: terminator (standard SSE).

$	curl -X POST https://api.empiriolabs.ai/v1/audio/speech:stream \
>	-H "Authorization: Bearer <token>" \
>	-H "Content-Type: application/json" \
>	-d '{
>	"model": "tts-1-5-mini",
>	"input": "Hello, this audio streams as it'\''s synthesized.",
>	"voice": "Sarah",
>	"output_format": "PCM",
>	"sample_rate": "24000",
>	"timestamp_type": "WORD"
>	}'