ElevenLabs API Python Quickstart: 5 Steps (2026)

The ElevenLabs API Python quickstart is five steps: install the SDK, create an API key, set the key, pick a voice, and call the text-to-speech endpoint. Every call spends credits at the standard 1-credit-per-character rate from your monthly pool. Checked 4 Nov 2026.

What do you need before you start?

Python 3.8 or newer, pip, and an ElevenLabs account. The SDK installs from PyPI in one command, and the free tier’s 10,000 monthly credits cover all the testing in this guide — no paid plan is required to follow along.

You also need an API key, which you create inside your ElevenLabs account settings. Keys are per-account secrets: never commit one to a public repository. Our ElevenLabs API key guide walks through creating and storing keys safely.

ElevenLabs API Python quickstart in 5 steps

Five steps from zero to generated audio. Credit costs are noted per step:

1. Install the SDK with pip install elevenlabs. Cost: 0 credits.
2. Create an API key in your ElevenLabs account and keep it private — store it as an environment variable, not in your code. Cost: 0 credits.
3. Initialize the client with your key. Cost: 0 credits.
4. Pick a voice ID from the Voices section of the app. Cost: 0 credits.
5. Call the text-to-speech endpoint and save the audio. Cost: 1 credit per character — “Hello world” (11 characters) costs 11 credits.

Steps 3 and 5 in code (method names follow the official SDK — check the docs for your installed version):

from elevenlabs.client import ElevenLabs
client = ElevenLabs(api_key=”YOUR_API_KEY”)

audio = client.text_to_speech.convert(
    voice_id=”YOUR_VOICE_ID”,
    text=”Hello world”,
    model_id=”eleven_multilingual_v2″,
)
with open(“hello.mp3”, “wb”) as f:
    for chunk in audio:
        f.write(chunk)

Only step 5 spends credits. Installing, key setup, and client initialization are free — the meter runs on the characters you send, so the five steps above cost 11 credits total for the “Hello world” test.

How to batch-generate hundreds of files

The quickstart covers one call. Production usually means hundreds — a course with 200 lessons, or a podcast feed with daily episodes. The pattern is a loop with a short pause between calls:

texts = [“Lesson 1 …”, “Lesson 2 …”]
for i, text in enumerate(texts):
    audio = client.text_to_speech.convert(
        voice_id=”YOUR_VOICE_ID”,
        text=text,
        model_id=”eleven_multilingual_v2″,
    )
    with open(f”lesson_{i}.mp3″, “wb”) as f:
        for chunk in audio:
            f.write(chunk)
    time.sleep(1)  # be polite to the API

Two rules keep batches healthy. First, the pause: rapid-fire requests can hit rate limits, and a one-second sleep between calls avoids most of them. Second, the budget check: 200 lessons at 3,000 characters each need 600,000 credits — exactly Pro’s monthly allowance — so price the batch before you run it.

Log every call’s character count as you go. A running total tells you when you are approaching the plan’s limit, and the log doubles as the backup record if you bill the generation work to a client.

API rate limits and retries

The API throttles aggressive callers. Fire dozens of requests per second and expect 429 responses — the server’s way of saying slow down. The fix is client-side: catch the error, wait a few seconds with exponential backoff, and retry the failed call.

Design for it from the start. Wrap the convert call in a retry loop (three attempts, doubling the wait each time), and treat a 429 like any other transient error — not a bug in your code. Failed calls never spend credits, so retries are free until they succeed.

How much does the API cost in credits?

Exactly the same as the web studio: one credit per character of input text. The API adds no surcharge and no per-call fee — a 5,000-character request costs 5,000 credits whether it comes from Python or the browser.

That makes budgeting straightforward. A script that generates one 2,000-character voiceover per day costs 60,000 credits a month — half of Creator’s 121,000. Ten such voiceovers a day needs 600,000, which is exactly Pro’s allowance. See our ElevenLabs API pricing guide for the full rate picture.

Test calls are cheap. The five steps above spend 11 credits total, so the free tier’s 10,000 monthly credits leave room for roughly nine hundred test runs before you spend a cent.

API vs web studio: when to use which

Use the API when generation must happen without a human clicking: scheduled content, user-facing app features, batch processing hundreds of scripts, or pipelines that chain transcription to voiceover.

Use the web studio when a human is in the loop: auditioning voices, tuning the stability and similarity sliders, or producing one-off audio. Our text to speech settings guide covers the slider side of that work.

Many teams use both: prototype the voice and settings in the studio, then hard-code the winning voice ID and settings into the Python script for production.

Common API errors and fixes

Invalid API key is the most common first error. It means the key string is wrong, revoked, or pasted with extra whitespace — regenerate the key in account settings and try again. The failed call costs 0 credits.

Insufficient credits means the account’s monthly pool is empty. Failed calls do not spend credits, but nothing generates until the next refill or an upgrade. A quota error mid-batch means the same thing — check the balance before retrying the batch.

Voice not found means the voice ID is mistyped or belongs to a different account. Voice IDs are long strings — copy them from the app rather than typing them, and confirm the voice exists on the key’s account.

Model not found means the model_id string is wrong or retired. Model names change as ElevenLabs ships new generations — copy the current model ID from the official docs rather than from an old tutorial, and pin the model in production code so an update never silently changes your voice.

ElevenLabs API Python: pros and cons

Who should NOT use the API

If you generate a handful of voiceovers a month by hand, the API buys you nothing — the web studio is faster and visual. The API pays off when generation is frequent, scheduled, or triggered by other software.

And if you cannot keep a secret safe, do not start. An API key in a public repo or a shared notebook will be found and spent. Learn key hygiene first — our API key guide covers the basics.

Frequently asked questions

Do I need a paid plan for the ElevenLabs API?

No. The API works on the free tier’s 10,000 monthly credits, which cover development and testing. Upgrade when your volume exceeds the free allowance — Starter’s 30,000 credits ($6/month) are the first paid step for API work. Choose the plan by monthly character volume.

Where do I find my ElevenLabs API key?

In your ElevenLabs account settings, under the API section. Create a key, copy it once, and store it as an environment variable — never hard-code it into shared files. See our API key guide for the full walkthrough linked above.

How many credits does one API call cost?

One credit per character of input text — the same rate as the web studio, with no per-call fee. A 2,000-character request costs 2,000 credits; divide your monthly allowance by your average request size to plan capacity before you build.

Can I use the API on the free tier?

Yes. Every endpoint in this quickstart works on the free 10,000-credit tier, including text-to-speech. The free tier’s non-commercial license applies to API output too — upgrade to a paid plan for commercial use on any endpoint. The license follows the plan, not the interface.

What Python version works with the ElevenLabs SDK?

Python 3.8 or newer. Install with pip install elevenlabs, then check the official docs for your SDK version’s method names, since the SDK is updated regularly and names occasionally change. Pin the version in requirements for production to avoid surprises.

How do I save API audio to a file?

The convert call returns audio chunks — write them to a file in binary mode, as shown in step 5. Name the file with an .mp3 extension and it plays in any standard player. The same pattern works for longer generations.

Sources checked

SDK usage and the 1-credit-per-character rate come from the official ElevenLabs documentation, checked 4 Nov 2026. Plan prices and credit allowances come from the official ElevenLabs pricing page, checked 4 Nov 2026.