Skip to main content
POST
/
tts
/
bytes
Text to Speech (Bytes)
curl --request POST \
  --url https://api.cartesia.ai/tts/bytes \
  --header 'Authorization: Bearer <token>' \
  --header 'Cartesia-Version: <cartesia-version>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model_id": "<string>",
  "transcript": "<string>",
  "voice": {
    "id": "<string>"
  },
  "output_format": {},
  "pronunciation_dict_id": "<string>",
  "generation_config": {
    "volume": 1,
    "speed": 1
  },
  "speed": "normal"
}
'
"<string>"

Documentation Index

Fetch the complete documentation index at: https://docs.cartesia.ai/llms.txt

Use this file to discover all available pages before exploring further.

Authorizations

​
Authorization
string
header
required

A short-lived access token to make API requests from a client.

Headers

​
Cartesia-Version
enum<string>
required

API version header.

Available options:
2026-03-01
Example:

"2026-03-01"

Body

application/json
​
model_id
string
required

The ID of the model to use for the generation. See Models for available models.

​
transcript
string
required
​
voice
TTSRequestVoiceSpecifier · object
required
​
output_format
RAWOutputFormat · object
required
​
language
enum<string> | null

The language that the given voice should speak the transcript in. For valid options, see Models.

Available options:
en,
fr,
de,
es,
pt,
zh,
ja,
hi,
it,
ko,
nl,
pl,
ru,
sv,
tr,
tl,
bg,
ro,
ar,
cs,
el,
fi,
hr,
ms,
sk,
da,
ta,
uk,
hu,
no,
vi,
bn,
th,
he,
ka,
id,
te,
gu,
kn,
ml,
mr,
pa
​
pronunciation_dict_id
string | null

The ID of a pronunciation dictionary to use for the generation. Pronunciation dictionaries are supported by sonic-3 models and newer.

​
generation_config
GenerationConfig · object

Configure the various attributes of the generated speech. Available on sonic-3 and sonic-3.5; not available on earlier models.

See Volume, Speed, and Emotion for a guide on this option.

​
speed
enum<string>
default:normal
deprecated

Use generation_config.speed for sonic-3. Speed setting for the model. Defaults to normal. This feature is experimental and may not work for all voices. Influences the speed of the generated speech. Faster speeds may reduce hallucination rate.

Available options:
slow,
normal,
fast

Response

200 - audio/*

Audio bytes

The response is of type file.