Sonic 3.5

Real-time text-to-speech model, built on State Space Models (SSMs) for streaming speech generation.

Model details

Developed by
Cartesia
Use case
text to speech

Example usage

Sonic 3.5 is Cartesia's fastest and most natural text-to-speech model, ranked #1 for naturalness with sub-90ms latency and native support for 42 languages. It follows transcripts faithfully, voices confirmation codes and heteronyms correctly without preprocessing, and stays expressive enough to carry a real conversation.

See full documentation

Input

1import requests
2
3url = "https://api.cartesia.ai/tts/sse"
4
5payload = {
6    "model_id": "sonic-3.5",
7    "transcript": "Hi there, it's awesome to meet you.",
8    "voice": {
9        "mode": "id",
10        "id": "db6b0ed5-d5d3-463d-ae85-518a07d3c2b4"
11    },
12    "output_format": { "container": "raw" },
13    "add_timestamps": False,
14    "add_phoneme_timestamps": False,
15    "use_normalized_timestamps": True,
16    "pronunciation_dict_id": "<string>",
17    "generation_config": {
18        "volume": 1,
19        "speed": 1
20    },
21    "speed": "normal",
22    "context_id": "<string>"
23}
24headers = {
25    "Cartesia-Version": "2026-03-01",
26    "Authorization": "$CARTESIA_API_KEY",
27    "Content-Type": "application/json"
28}
29
30response = requests.post(url, json=payload, headers=headers)
31
32print(response.text)

JSON output

1{
2    "type": "chunk",
3    "done": false,
4    "status_code": 206,
5    "step_time": 123,
6    "context_id": "50dc3b5e-5841-4aa1-9f94-60cfb9aead79",
7    "data": "aSDinaTvuI8gbWludGxpZnk="
8}