Text to speech, paid per call

Send up to 3,000 characters and a voice, settle the 402 in USDC, and an ElevenLabs MP3 comes back in seconds. Priced by the character, quoted before you pay.

POST /api/x402/speech · from $0.02

Turn the sound on: the narration is ElevenLabs, from one of our own channel's Shorts. It was made on our pipeline, not bought here; the speech endpoint sells the same voices, the MP3 alone.ElevenLabs

Try it

1 · Ask for the price (free)

curl -i -X POST https://www.trezalabs.com/api/x402/speech \
  -H 'Content-Type: application/json' \
  -d '{"text":"Welcome back. Today we are looking at three small habits that make a big difference.","voice":"george","model":"eleven-v3"}'

Without a payment this answers 402 Payment Required, with the price for exactly this body, the USDC token and the address to pay in the PAYMENT-REQUIRED header. Nothing is charged.

2 · Pay and collect, with Coinbase's awal wallet

npx awal@latest x402 pay https://www.trezalabs.com/api/x402/speech \
  -X POST \
  -d '{"text":"Welcome back. Today we are looking at three small habits that make a big difference.","voice":"george","model":"eleven-v3"}' \
  --max-amount 40000

awal reads the 402, pays it from your agentic wallet in USDC on Base, repeats the request and prints the response. --max-amount caps the payment in USDC units of a millionth: 40,000 is $0.04, this body's price. Any other x402 client works the same way.

Or just read what it asks for

npx awal@latest x402 details https://www.trezalabs.com/api/x402/speech

Price

Quoted per request in the 402, so read it from there rather than hard-coding it. A render that fails gets a free retry.

ElevenLabs Eleven v3defaulteleven-v3
$0.42 per 1,000 characters, minimum $0.02
ElevenLabs Multilingual v2eleven-multilingual-v2
$0.42 per 1,000 characters, minimum $0.02

Limits

text: up to 3,000 characters.
Returns an MP3 URL in the JSON response, with a statusUrl you can GET again for free.
60 requests per hour per IP address, before any payment is looked at.
Paid in USDC on Base or Solana. Fields the endpoint does not read are refused before payment, not charged for and ignored.
Eleven v3 takes delivery directions in square brackets, like [whispers] or [excited], and performs them without reading them out. Multilingual v2 reads 29 languages.
A voice whose owner charges more carries a priceMultiplier on the menu, and the 402 quotes it.

A bare GET /api/x402/speech answers with the whole menu as JSON.

What a paid call returns

{
  "runId": "2026-09-23T18:04:11.000Z#1a2b3c4d",
  "status": "success",
  "audio": "https://www.trezalabs.com/api/media/generated/x402_speech/1a2b3c4d.mp3?s=...&c=...",
  "statusUrl": "https://www.trezalabs.com/api/x402/speech?runId=...&token=...",
  "paidUsd": 0.04,
  "model": "eleven-v3",
  "voice": "george",
  "characters": 84,
  "transaction": "0x..."
}

The same payment buys

Video

POST /api/x402/video

A prompt or a first frame in, a clip out, on Veo 3.1, Kling 3.0, Seedance 2.5, Wan 2.7 or Hailuo 3.

From $0.42

Music

POST /api/x402/music

An original Google Lyria 3 track from a description: a 30-second bed or a full song.

From $0.06

Images

POST /api/x402/image

Nano Banana, GPT Image 2, Seedream 5.0 Pro, FLUX.2 Pro and Recraft, up to 4K.

From $0.02

Shorts

POST /api/x402/short

A topic in, a finished 30-second narrated vertical short out: scenes, voice, music, captions.

From $0.87

Listed in the Coinbase x402 Bazaar, so agents can find it with npx awal@latest x402 bazaar search "text to speech". Full reference in the x402 docs and the OpenAPI spec.