BlockRun
ServicesBy capability: Text to Speech API

PARTNER

ElevenLabs Voice

The most realistic AI voice, priced per call and quoted before it runs.

Give your agent a voice. ElevenLabs text-to-speech and sound effects through one endpoint — Flash v2.5 for real-time conversation, Multilingual v2 / Eleven v3 for studio-grade narration. Price is quoted up front, you pay only after the audio is generated, and there's no subscription to manage.

HOW IT WORKS

Send text to /v1/audio/speech; the server returns a 402 with the exact price (characters × model rate). Sign the payment, replay the request, and get back a hosted audio URL. Flash starts at $0.0535/1k chars, minimum $0.002 per request (a $0.001 floor plus the $0.001 per-request fee).

3 endpoints live on Base
32 languages
~75ms latency (Flash)
No account required

What Agents Use It For

Anywhere your agent needs to speak — voice assistants, narration, IVR, accessibility, game and video audio — without standing up a TTS pipeline or committing to a monthly plan.

Real-Time Agent Voice
Flash v2.5 synthesizes speech at ~75ms latency — the voice your agent speaks with in real time. 32 languages, deterministic per-character pricing.
Studio-Grade Narration
Multilingual v2 and Eleven v3 deliver maximum consistency and emotional range for audiobooks, voiceover, and long-form content.
Cinematic Sound Effects
Generate sound effects and audio textures from a text prompt — explosions, ambience, UI sounds, foley — up to 22 seconds per clip.
Pay-per-Call, No Accounts
Every call settled in USDC via x402 on Base. No ElevenLabs subscription, no API keys for callers — just attach payment to the request.

Models & Pricing

Text-to-speech is billed per input character (quoted before you pay). Sound effects are flat per generation. Prices below are what you pay for 1,000 characters or one clip — the model rate, the 5% platform fee and the $0.001 per-request fee included — with a $0.002 per-request minimum.

ModelPriceMax inputBest for
elevenlabs/flash-v2.5$0.0535 / 1k chars40,000Ultra-low-latency (~75ms) speech synthesis for real-time voice agents. 32 languages.
elevenlabs/turbo-v2.5$0.0535 / 1k chars40,000Balanced quality and latency (~250ms) for interactive use cases. 32 languages.
elevenlabs/multilingual-v2$0.106 / 1k chars10,000Highest-consistency voice for long-form narration, audiobooks, and voiceover. 29 languages.
elevenlabs/v3$0.106 / 1k chars5,000Maximum expressiveness and emotional range for creative applications. 70+ languages.
bytedance/seed-audio-1.0$0.316 / 1k chars3,000ByteDance's Seed Audio 1.0 — prompt-directed audio creation: describe the voice, emotion, and sound staging in natural language. Up to 120s output, mp3/wav. Billed by audio duration ($0.003/second, estimated from input length).
elevenlabs/sound-effects$0.0535 / clip22sGenerate cinematic sound effects and audio textures from a text prompt (up to 22s).

Quick Start

01

Pick a voice (free)

curl https://blockrun.ai/api/v1/audio/voices
# -> { "data": [ { "voice_id": "EXAVITQu4vr4xnSDxMaL", "alias": "sarah", ... }, ... ] }
02

Synthesize speech (pay per call)

Send unpaid to get a 402 with the exact price, then replay with the signed x-payment header.

curl -X POST https://blockrun.ai/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
    "model": "elevenlabs/flash-v2.5",
    "input": "Hello from BlockRun. Pay per call, no subscription.",
    "voice": "sarah",
    "response_format": "mp3"
  }'

# Response (after settlement):
# {
#   "model": "elevenlabs/flash-v2.5",
#   "data": [{ "url": "https://blockrun.ai/api/media/media/audios/...mp3",
#              "format": "mp3", "characters": 51 }]
# }

Use From Your Tooling

Claude Code / Cursor (MCP)

Install once, then ask: "use blockrun to speak this with the sarah voice."

claude mcp add blockrun -s user -- \
  npx -y @blockrun/mcp@latest
Python / TypeScript SDK

Wallet loaded from ~/.blockrun/.session. SDK handles 402 + signing.

pip install blockrun-llm
# or
npm install @blockrun/llm

Behind the Scenes

ElevenLabs

Powers the voice synthesis and sound generation. BlockRun holds the ElevenLabs API key server-side and covers the upstream cost — you pay per call in USDC, with no subscription.

elevenlabs.io

Get Started

All you need is a USDC-funded wallet on Base. Hit any endpoint, get a 402, attach payment, get the audio.

ElevenLabs API questions: voices, languages, pricing, sound effects

Which ElevenLabs models are available?
The models on this page's rate card: the flash tier for low-latency conversation, the multilingual and v3 tiers for expressive long-form narration, and the sound-effects model on its own endpoint.
What languages do the ElevenLabs voices speak?
The multilingual models cover dozens of languages; the voices endpoint lists which voices speak which, and the model's row links to the full list.
What is the latency of ElevenLabs Flash through BlockRun?
Flash is the real-time tier and the fastest on the card; the figure on this page is what it measures. The other tiers trade latency for expressiveness.
Can I generate sound effects with the ElevenLabs API here?
Yes — a text description on the sound-effects endpoint returns a short clip, priced per clip, on the same payment flow.
Can I clone a voice through this endpoint?
No. Voice cloning is an ElevenLabs console feature; this endpoint synthesises speech from the voices the models already offer.
ElevenLabs through BlockRun vs ElevenLabs direct, OpenAI TTS or Google speech?
Direct is the same voices behind an ElevenLabs subscription and account; OpenAI and Google sell their own voices behind theirs. Here the same ElevenLabs models are paid per request with no account, beside other vendors' voices on one endpoint.
How do I use the ElevenLabs API without an ElevenLabs subscription?
Call BlockRun's speech endpoint. It carries ElevenLabs' voice models — Flash for real-time, Multilingual and Eleven v3 for studio narration — priced per thousand characters and paid per request, with no ElevenLabs account, seat or monthly allowance.
How much does ElevenLabs text-to-speech cost per call?
The per-thousand-character rate on each model's row, times the characters you send, plus the flat per-call fee — quoted in the 402 before the audio is synthesised.
Can an AI voice agent call ElevenLabs without a human account?
Yes. Each request pays for itself, so a voice agent with a funded balance speaks with no signup and nothing to cancel.