Cartesia Sonic

Overview

Best for: Latency-critical voice agents and telephony (native mu-law 8 kHz), with timestamps for interruption handling.

At a glance

Price is the Pro plan effective rate ($5 for 100K credits, 1 credit per char); Pro overage $65/1M, Scale about $37/1M. Latency ~90 ms is a vendor claim quoted by third parties. Concurrency 3 on Pro (Free 2, Scale 15). Break tags supported but not full SSML. Free plan not listed for commercial use.

Features

• input streaming via contexts (continue flag)
• word timestamps
• phoneme timestamps
• emotion, speed and volume controls (generation_config)
• voice cloning
• pronunciation dictionaries
• cancel per context
• access tokens for browser clients

Pricing

Free tier: 20K credits/month, overages blocked; commercial use not listed for Free (Pro and up: yes).

Source: cartesia.ai

Setup

1. Create an API key in the Cartesia playground/console.
2. Pick a voice ID.
3. Open wss://api.cartesia.ai/tts/websocket with X-API-Key and a cartesia_version.
4. Send generation requests sharing one context_id; set continue:true on every chunk except the last.
5. Read base64 'chunk' messages until 'done'.

Endpoint

wss://api.cartesia.ai/tts/websocket

Authentication

X-API-Key header (server) or access_token query param (browser); cartesia_version required (example 2026-08-14)

# pip install websockets   (streams text chunks into one context)
import asyncio, base64, json, os, websockets

URL = "wss://api.cartesia.ai/tts/websocket?cartesia_version=2026-08-14"
HDR = {"X-API-Key": os.environ["CARTESIA_API_KEY"]}

async def main():
    async with websockets.connect(URL, additional_headers=HDR) as ws:
        chunks = ["Hello there. ", "This arrives ", "token by token."]
        for i, text in enumerate(chunks):
            await ws.send(json.dumps({
                "model_id": "sonic-3.6",
                "transcript": text,
                "voice": {"mode": "id", "id": "YOUR_VOICE_ID"},
                "output_format": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 24000},
                "context_id": "turn-1",
                "continue": i < len(chunks) - 1,
            }))
        with open("out_24k_s16le.pcm", "wb") as f:
            async for raw in ws:
                msg = json.loads(raw)
                if msg.get("type") == "chunk":
                    f.write(base64.b64decode(msg["data"]))
                elif msg.get("type") in ("done", "error"):
                    break

asyncio.run(main())

Written from the current docs. Check the vendor's SDK version before you ship.

Warnings

Plus 12 warnings that apply to all text-to-speech, streaming APIs. See category warnings.

Limits

• TTS concurrency: Free 2, Pro 3, Startup 5, Scale 15
• max_buffer_delay_ms default 3000 (0-5000)
• voice and output_format must stay constant within a context
• Overages must be enabled; Free stops at the credit cap

Models and products

Docs and sources

Docs

• Models
• WebSocket API
• Pricing

Sources used

• cartesia.ai/pricing
• docs.cartesia.ai/build-with-cartesia/tts-models/latest
• docs.cartesia.ai/api-reference/tts/websocket
• humannessindex.vapi.ai/models/cartesia-sonic-3-5