Stream speech with word timestamps

Stream speech and word timestamps as newline-delimited JSON (NDJSON). Each line contains audio, timing updates, or an error. See Speech timing for stream handling.

POST /v1/text-to-speech/{voice_id}/stream/with-timestamps

cURL

cURL

curl \
  --request POST \
  --url "https://api.breeze.blue/v1/text-to-speech/voc_xeh3w54cqvnp/stream/with-timestamps" \
  --header "xi-api-key: $BREEZE_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "text": "Hello world.",
  "model_id": "breeze-tts-2-multilingual",
  "language_code": "en",
  "timestamp_mode": "chunk"
}'

200

{
  "audio_base64": "<string>",
  "word_timestamps": [
    {
      "index": 1,
      "word": "<string>",
      "start": 1,
      "end": 1
    }
  ]
}

Continue building

Write instructions

Choose intent, emotion, and delivery.

Speech timing

Build captions with word timing.

Convert with timestamps

Get complete audio with word timing.

Output formats

Choose and decode an audio format.

Authorizations

xi-api-key · string · header · required

Breeze Developer API key.

Path Parameters

voice_id · string · required

Voice ID. See List voices.

Query Parameters

output_format · string | null · default: pcm

Audio encoding: pcm, mp3, or wav. Default: pcm. Decode audio_base64 before playing. See Output formats.

Body

application/json

text · string · required

Text to synthesize. Up to 1000 characters by default; accounts with an approved higher limit may send up to their configured limit, at most 2000 characters. See Audio tags.

Minimum string length: 1

model_id · string | null

Model ID for speech generation. Selected automatically when omitted. See List models.

Required string length: 1 - 120

language_code · string | null

ISO 639-1 two-letter language code supported by the selected model. See supported language codes.

Required string length: 2

Pattern: ^[A-Za-z]{2}$

instructions · string | null

Performance instructions, written in the same language as the input text. See Expressive controls, Voice instruction prompting.

voice_settings · TtsVoiceSettingsPayload · object | null

Optional per-request voice settings override. See Voice settings.

Show child attributes

timestamp_mode · enum<string> · default: chunk

See Streaming timing modes.

Available options: chunk, lookahead

Response

200

application/x-ndjson

NDJSON; each line is a speech timing chunk. A stream error instead contains an error object with code and message.

audio_base64 · string · required

Base64 audio bytes; may be empty for metadata-only updates. See Decode audio and timing.

word_timestamps · WordTimestampResponse · object[] · required

See Speech timing.

Show child attributes