ElevenLabs v3 - AI Model API and Pricing

Status: live in two tools

The rate is 15 credits per 1,000 characters. The AI Voice Generator starts on a livelier read for ads and voiceovers, and the AI Narrator on a steady one for lessons and documentaries. Both offer the same stock voices, 14 languages, three speeds and three levels of expression, and take up to 3,000 characters per run.

What it is built for

ElevenLabs built v3 for expressive speech: voiceovers, ads and narration that need emotion, in more than 70 languages. Its listed weak points are that it needs more script work than earlier ElevenLabs models and that it is not meant for real-time conversation. On Ugesi you pick one of 14 languages. The setting tells the voice how to read the text and does not translate it.

The main control is the script itself. Audio tags in square brackets, such as [whispers], [excited], [sighs] or [laughs], change how the next words are read. Example: "[whispers] The results are in. [excited] And we beat last year!" Use one tag per sentence at most, and expect to generate a key line twice.

Expression and stability

The Expression setting maps to the model's stability. Steady is the most stable read and follows tags least. Natural sits in the middle. Expressive gives the widest emotional range and the most variation between runs. In a long narration, Steady keeps every section sounding alike. In a short ad, Expressive adds energy.

Budgeting a script

1,000 characters is about 150 spoken English words, roughly a minute, and costs 15 credits, about $0.16. Each started 1,000 counts in full, so 1,200 characters cost 30 credits. A full 3,000-character run costs 45 credits, about $0.48. It ranks 5th from the cheapest of 6 comparable speech models, which run from 5 to 16 credits per 1,000 characters.

Neighboring voice models

IndexTTS 2, at 7 credits, is aimed at everyday voiceover in volume. XTTS v2, at 10 credits, reads one script in many languages. VibeVoice, at 6 credits, handles long dialogue with several speakers. None of them is in a site tool today. Check each model page for its current status.

Where to actually run it

ElevenLabs v3 API

curl -X POST https://ugesi.com/api/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "elevenlabs-v3",
    "prompt": "Welcome to our studio. Here is what is new this week."
  }'

<?php
$ch = curl_init('https://ugesi.com/api/generate');
curl_setopt_array($ch, [
    CURLOPT_POST => true,
    CURLOPT_RETURNTRANSFER => true,
    CURLOPT_HTTPHEADER => [
        'Authorization: Bearer YOUR_API_KEY',
        'Content-Type: application/json',
    ],
    CURLOPT_POSTFIELDS => json_encode([
        'model' => 'elevenlabs-v3',
        'prompt' => 'Welcome to our studio. Here is what is new this week.',
    ]),
]);
$response = json_decode(curl_exec($ch), true);
curl_close($ch);
print_r($response);

type GenerateRequest = {
  model: string;
  prompt: string;
};
type GenerateResponse = {
  job_id: number;
  job_token: string;
  credits_held: number;
  status: "queued" | "succeeded";
  output_text?: string;
};

const body: GenerateRequest = {
  model: "elevenlabs-v3",
  prompt: "Welcome to our studio. Here is what is new this week.",
};

const resp = await fetch("https://ugesi.com/api/generate", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.UGESI_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify(body),
});
if (!resp.ok) throw new Error(`HTTP ${resp.status}`);
const job = (await resp.json()) as GenerateResponse;
console.log(job.job_id, job.status);

import requests

resp = requests.post(
    "https://ugesi.com/api/generate",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "elevenlabs-v3",
        "prompt": "Welcome to our studio. Here is what is new this week.",
    },
    timeout=60,
)
resp.raise_for_status()
print(resp.json())

{
  "model": "elevenlabs-v3",
  "prompt": "Welcome to our studio. Here is what is new this week."
}

The API comes with every credit pack, from Lite to Enterprise, and uses the same balance as the website. API documentation and access →